Skip to content

Claude/confirm doc context loading r0 pa5#1583

Closed
Censseo wants to merge 92 commits intogithub:mainfrom
Censseo:claude/confirm-doc-context-loading-R0PA5
Closed

Claude/confirm doc context loading r0 pa5#1583
Censseo wants to merge 92 commits intogithub:mainfrom
Censseo:claude/confirm-doc-context-loading-R0PA5

Conversation

@Censseo
Copy link
Copy Markdown

@Censseo Censseo commented Feb 8, 2026

No description provided.

Censseo and others added 30 commits November 17, 2025 18:10
…oks configuration

Add new /speckit.setup-hooks command that:
- Detects project type (Node, Python, Rust, Go, Java, .NET, Ruby, PHP)
- Identifies package managers, test runners, linters, and formatters
- Generates .claude/settings.json with SessionStart, PostToolUse, and Stop hooks
- Creates skill templates for testing and linting
- Provides cross-platform support (bash and PowerShell scripts)

This enables automated project setup for Claude Code with:
- Automatic dependency installation on session start
- Auto-formatting after file edits
- Test execution before session end
…itecture skills

Enhance the setup-hooks command with:

- Scala project detection (sbt and Mill build tools)
- Framework detection for 50+ frameworks across all languages:
  - Frontend: React, Vue, Angular, Svelte, Next.js, Nuxt, Astro, Remix
  - Backend: Express, FastAPI, Django, Spring, NestJS, Gin, Axum
  - Scala: Play, Akka, ZIO, Cats Effect, http4s, Tapir
  - And many more...
- Documentation URLs for each framework (docs + GitHub)
- Architecture skills generation from framework documentation
- Language-specific coding standards skills (TypeScript, Python, Scala)
- WebFetch integration to retrieve best practices from official docs

The command now generates comprehensive skills that include:
- Project structure recommendations
- Core architectural patterns
- Best practices (do's and don'ts)
- Coding conventions
- Performance guidelines
- Security considerations
…ices

- Add references/ folder structure for detailed docs (patterns, best-practices, examples)
- Keep SKILL.md lean with quick reference tables
- Add trigger phrases in skill descriptions for better auto-activation
- Explain when to load references for context efficiency
- Update language-specific skills (TypeScript, Python, Scala, Rust, Go)
- Improve completion report with new folder structure
Major refactoring of /agents command to work with skills:

- Add SpecKit SDD workflow agents: spec-analyzer, designer, implementer, tester
- Agents now reference skills via `skills:` frontmatter field
- Add official frontmatter format: name, description, tools, model, skills
- Add model selection guide (haiku for fast tasks, sonnet for complex)
- Update discovery flow to find existing skills from /setup-hooks
- Add architecture diagram showing skills + agents relationship
- Keep agents lean (orchestration) while skills provide expertise
- Document /setup-hooks → /agents workflow

Skills provide domain expertise; Agents orchestrate workflows.
- Add "When to Use Opus" section with specific scenarios
- Add orchestrator logic for dynamic model selection
- Add cost/speed tradeoffs table
- Document when to promote from sonnet to opus:
  - Architecture decisions (10+ components)
  - Security-critical features
  - Major refactoring
  - Cross-system integration
Merged changes from upstream:
- Add Qoder CLI and IBM Bob as supported agents
- Fix global branch numbering to prevent collisions
- Fix octal interpretation in branch numbers (010 → 10 not 8)
- Update quickstart guide and documentation
- Formatting improvements

Note: Excluded .github/workflows changes due to permissions
- Add frontend-designer agent between designer and implementer
- Uses Anthropic's frontend-design skill for distinctive aesthetics
- Focus on avoiding generic AI aesthetics (purple gradients, Inter font, etc.)
- Outputs design specs: typography, colors, layout, interactions
- Update workflow diagrams to show frontend-designer branch
- Add to model selection guide (sonnet for creative reasoning)

Workflow: designer → frontend-designer → frontend-coder
Problems fixed:
- Subagents not being invoked → Added CRITICAL RULES section with MUST language
- tasks.md not updated → Rule 2: ALWAYS update immediately after each task
- Context bloat → Rule 4: Minimal context loading, agents load their own

Changes:
- Add 4 CRITICAL RULES at top of command (read first)
- Simplify agent selection with clear pattern→agent table
- Restructure task execution as 5-step loop (7.1-7.5)
- Add explicit Task tool invocation template
- Add immediate post-task checklist (create result, update tasks.md, log)
- Add pause checkpoints every 3 tasks
- Reduce upfront context loading (tasks.md + plan.md only)
fix(implement): enforce agent invocation and immediate updates

Problems fixed:
- Subagents not being invoked → Added CRITICAL RULES section with MUST language
- tasks.md not updated → Rule 2: ALWAYS update immediately after each task
- Context bloat → Rule 4: Minimal context loading, agents load their own

Changes:
- Add 4 CRITICAL RULES at top of command (read first)
- Simplify agent selection with clear pattern→agent table
- Restructure task execution as 5-step loop (7.1-7.5)
- Add explicit Task tool invocation template
- Add immediate post-task checklist (create result, update tasks.md, log)
- Add pause checkpoints every 3 tasks
- Reduce upfront context loading (tasks.md + plan.md only)
- Add templates, scripts, and memory to pip package via hatch force-include
- Add build_template_from_bundled() to construct templates from bundled files
- Modify download_and_extract_template() to use bundled templates first
- Add --force-download flag to force GitHub download instead of bundled
- Bump version to 0.0.23

This allows users installing from a fork to get the fork's templates
instead of downloading from upstream github/spec-kit releases.
New command that runs the complete initial setup sequence:
1. /speckit.setup-hooks - Detect frameworks, create skills
2. /speckit.agents - Configure project agents
3. /speckit.constitution - Establish project principles

This simplifies onboarding by running all setup steps with a single command.
The previous version tried to invoke other slash commands which doesn't
work in Claude Code. Now the setup command contains all logic inline:

- Phase 1: Run detection script, create skills, configure hooks
- Phase 2: Create workflow agents (researcher, planner, coders, tester)
- Phase 3: Initialize project constitution

All actions are performed directly instead of delegating to sub-commands.
fix(commands): rewrite /speckit.setup to execute actions directly
- Add /speckit.review command for code quality analysis:
  - Pre-implementation review (spec alignment)
  - Post-implementation review (quality gates)
  - Full audit mode with tech debt scoring
  - Focused reviews on specific areas

- Add code-review skill with references:
  - Comprehensive review checklist (architecture, quality, security, etc.)
  - OWASP Top 10 security review guide
  - Language-specific security checks

- Add tech-debt skill with references:
  - 14 common debt patterns (God Object, Spaghetti, etc.)
  - Refactoring strategies (Extract Method, Strangler Fig, etc.)
  - Prioritization framework

- Update /speckit.setup to:
  - Create universal skills (code-review, tech-debt) in Phase 1
  - Include reviewer agent in Phase 2 with code-review/tech-debt skills
- Rewrite setup.md to properly compile all 3 commands:
  - Phase 1: Full setup-hooks content (hooks, skills, framework detection)
  - Phase 2: Full agents content (9 agents, discovery, model selection)
  - Phase 3: Full constitution content (template filling, principles)

- Add comprehensive Java/Maven and Java/Gradle support:
  - Hook scripts for Maven (mvn) and Gradle (./gradlew)
  - Java standards skill with Spring patterns
  - Google Java Format integration

- Include all hook script templates (session-setup.sh, auto-format.sh, pre-commit-checks.sh)
- Include all 10 agents with full system prompts
- Include language-specific standards skills (TS, Python, Java, Scala, Rust, Go)
- Add Skills + Agents architecture diagram
- Bump version to 0.0.25
Add 3 new universal skills for architecture and planning:

- architecture-patterns: Clean Architecture, Hexagonal, DDD
  - references/implementations.md: Production code examples
  - references/ddd-patterns.md: Strategic and tactical DDD patterns

- microservices-patterns: Service decomposition, communication, resilience
  - references/resilience-patterns.md: Circuit breaker, retry, bulkhead
  - references/data-patterns.md: Saga, event sourcing, CQRS, outbox

- architecture-decision-records: ADR templates and process
  - references/templates.md: MADR, lightweight, Y-statement, RFC formats
  - references/examples.md: Real-world ADR examples

Update /speckit.setup:
- Add all 3 skills to universal skills section
- Add architecture skills to designer agent
- Add ADR skill to planner agent
- Update completion summary and file tree

Source: wshobson/agents GitHub repository
Add a new command that transforms raw ideas into structured vision
documents before formal specification. This reduces ambiguity and
provides richer context for /speckit.specify.

- Add idea.md command with interactive discovery workflow
- Add idea-template.md for consistent output structure
- Update specify.md to detect and use existing idea documents
- Enable natural workflow: idea → specify → clarify → plan
- Use diff against main/master when on feature branch (not just HEAD~5)
- Insert correction tasks after last completed task, not at file end
- Add impact analysis on pending tasks with amendments
- Add blocking dependencies between review and affected tasks
- Document insertion point logic based on implementation progress
Add automated integration testing capabilities:

- Add /speckit.mcp command to generate project-specific MCP servers
- Add /speckit.validate command to run acceptance scenario tests
- Add MCP server template with tools for:
  - Process management (start/stop services, logs, health checks)
  - Browser automation (Playwright-based UI testing)
  - API testing (HTTP client with assertions)
  - Docker Compose integration
- Update /speckit.setup with optional Phase 4 for MCP configuration

Workflow: implement → review → validate (automated) → fix issues
Extend browser automation with:

Navigation:
- goBack, goForward, reload
- waitForNavigation, waitForLoad

Multi-tab:
- newTab, switchTab, closeTab, listTabs

iFrames:
- frameClick, frameFill, frameGetText

File upload:
- uploadFile, uploadFiles

Storage:
- localStorage: get/set/clear
- sessionStorage: get/set
- cookies: get/set/clear

Network:
- getNetworkRequests, waitForRequest, waitForResponse
- mockRequest, clearMocks

Interactions:
- hover, doubleClick, rightClick
- dragAndDrop, scroll, focus
- check, uncheck, clear
- getAttribute, getBoundingBox
- count, getAllTexts

Viewport:
- setViewport, getViewport, emulateDevice
- handleDialog (alert/confirm/prompt)
- Add complexity analysis phase with scoring (0-10 scale)
- Generate feature files for moderate/complex ideas (score >= 4)
- Create features/ subdirectory with numbered feature documents
- Update idea.md template with Features Overview section
- Add feature-template.md for individual feature documents
- Update /speckit.specify to load features incrementally with idea context
- Support feature file paths and numbers as input to specify command
- Track feature status through specification workflow
… features

Adds a new command that handles the case where a feature is already
implemented but not working correctly. Instead of starting from scratch
with /speckit.idea, users can now diagnose and fix existing implementations.

The command:
- Analyzes gaps between spec and implementation
- Classifies issues (spec gap, implementation bug, misunderstanding)
- Generates targeted correction tasks (FIX-### and CLARIFY-###)
- Integrates with existing commands via handoffs

Also updates handoffs in implement.md, review.md, and validate.md
to reference the new fix command where appropriate.
…MC0Z3

feat(fix): add /speckit.fix command for diagnosing poorly implemented…
Censseo and others added 26 commits January 28, 2026 11:28
…Hz93X

refactor: make setup scripts agent-agnostic with --agent-dir parameter
The commit 5042c76 ("Template cleanup and reorganization") was too
aggressive and removed document structure sections that belong in
the templates, not just workflow sections.

Restored in spec-template.md:
- CONSTITUTION COMPLIANCE comment block
- Error Scenarios section (mandatory per constitution)
- Accessibility Requirements section (mandatory for UI features)
- Performance Requirements section (conditional)
- Security Considerations section (conditional)
- Data & State section (conditional)

Restored in plan-template.md:
- Progress Tracking section (Phase Status + Gate Status)

https://claude.ai/code/session_01HUMuWguW4wboozXDdqaj2d
opencode and codex use AGENTS.md (shared with amp, q, bob), not
OPENCODE.md/CODEX.md. Aligned AGENT_CONFIG with the mappings in
update-agent-context.sh.

https://claude.ai/code/session_01TEXVWs8ymoUSvpzpirr3er
The agents.md template now includes format specifications for both
Claude Code and opencode:

- Format A (Claude Code): Simple frontmatter with comma-separated tools/skills
- Format B (opencode): Explicit boolean tool flags, mode, temperature, permissions
- Format C: Other agents follow Claude Code format

Added conversion notes and tool/model mapping tables to help agents
generate correct frontmatter for their platform. Templates show Claude
Code format for brevity with conversion instructions for other platforms.

Fixes opencode "expected record, received string tools" error by
documenting the correct tools object format: `tools:\n  read: true`
instead of `tools: Read, Glob, Grep`.

https://claude.ai/code/session_01TEXVWs8ymoUSvpzpirr3er
Previously, the update-agent-context scripts checked if the ENTIRE tech
stack string existed in the file. Since each feature has slightly
different framework combinations (same base like Java 21 but different
dependencies), every feature added its own full tech stack entry,
causing massive duplication in the Active Technologies section.

Now the scripts:
- Split tech stack strings into individual technologies
- Extract existing technologies from the file
- Only add technologies that don't already exist
- Prevent duplicate entries across multiple features

This results in cleaner, deduplicated Active Technologies sections.

https://claude.ai/code/session_012P9z9cjWNb4L9x5ZNKM6qM
…-uvwYM

fix: deduplicate tech stack at individual technology level
…Hz93X

Claude/check agent compatibility hz93 x
fix(templates): restore sections lost during template cleanup
Claude Code and opencode have different MCP configuration:

- Claude Code: `.claude/mcp.json` with `mcpServers` key
- opencode: `opencode.json` in project root with `mcp` key

opencode format requires:
- `type: "local"` or `"remote"`
- `command` as array (not separate command/args)
- `enabled: true/false`

Updated setup.md and mcp.md templates to show both formats.

Sources:
- https://opencode.ai/docs/mcp-servers/
- https://opencode.ai/docs/config/

https://claude.ai/code/session_01TEXVWs8ymoUSvpzpirr3er
…Hz93X

docs: add opencode MCP configuration format to templates
The previous fix only prevented adding new duplicates but didn't clean
up existing ones. This change rewrites the "Active Technologies"
section entirely with a consolidated, deduplicated list.

Changes:
- Add consolidate_technologies function that merges all technologies
- Categorize into two groups: Languages/Frameworks and Databases
- Strip parenthetical annotations like "(primary)", "(sessions/cache)"
  to properly deduplicate database entries
- Replace all existing tech entries with consolidated list on each run
- Result: Instead of 30+ duplicated lines per feature, just 2 clean
  lines (one for langs/frameworks, one for databases)

Example transformation:
Before:
- Java 21, TypeScript 5.x + Spring Boot, React 18 (001-feature)
- PostgreSQL 16 (primary), Redis 7 (sessions) (001-feature)
- Java 21, TypeScript 5.x + Spring Data JPA (002-feature)
- PostgreSQL 16, Redis 7 (003-feature)

After:
- Java 21, React 18, Spring Boot, Spring Data JPA, TypeScript 5.x
- PostgreSQL 16, Redis 7

https://claude.ai/code/session_012P9z9cjWNb4L9x5ZNKM6qM
…-uvwYM

fix: consolidate tech stack to eliminate duplications
Major changes:
- Rename /speckit.extract-patterns to /speckit.learn
- Add /speckit.merge command to consolidate specs into /docs
- Update architecture-registry.md to HIGH LEVEL only (patterns, contracts)
- Add module CLAUDE.md template for local conventions (auto-loaded)
- Update specify.md and plan.md to load /docs for consistency

New workflow: specify → plan → tasks → implement → merge → learn
- /docs becomes source of truth for business specs
- Module CLAUDE.md files auto-loaded by Claude Code during implementation
- Architecture registry stays lean with cross-cutting patterns only

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
- /docs organized by domain (e.g., /docs/auth/, /docs/payments/)
- Domains are user-defined, not hardcoded
- merge.md infers domain from feature, confirms with user
- specify.md and plan.md load domain-specific context
- Delta-style merging: ADDED/MODIFIED/REMOVED sections

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
New command to initialize /docs structure from:
- from-code: Analyze codebase, discover domains from modules/routes/entities
- from-docs: Convert existing README/wiki to domain structure
- from-specs: Migrate legacy specs/{feature}/ to /docs/{domain}/

Features:
- Auto-detect project state (code, docs, specs)
- Suggest domain groupings with user confirmation
- Generate draft specs with source traceability
- Non-destructive: creates new files, never deletes

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
Before: 6+ setup commands (setup, setup-hooks, bootstrap, constitution, agents, mcp)
After: 3 commands with clear responsibilities

- /speckit.setup: Orchestrator that runs bootstrap then agents
- /speckit.setup-bootstrap: Constitution + /docs init + learn
- /speckit.setup-agents: Agents + skills + MCP server

Removed:
- setup-hooks.md (hooks not currently useful)
- bootstrap.md (merged into setup-bootstrap)

Kept as building blocks:
- constitution.md, agents.md, mcp.md (for standalone use)

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
Instead of duplicating logic, setup-agents orchestrates:
- /speckit.agents (agents + skills)
- /speckit.mcp (MCP server)

Keeps commands modular and reusable.

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
New structure:
- /speckit.setup → orchestrator (bootstrap + agents)
- /speckit.setup-bootstrap → orchestrator (constitution + docs + learn)
  - /speckit.setup-constitution → calls constitution.md
  - /speckit.setup-docs → init /docs (from-code/from-docs/from-specs)
- /speckit.setup-agents → orchestrator (agents + mcp)
  - /speckit.agents
  - /speckit.mcp

Each sub-command can be run individually or via orchestrators.

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
Structure:
/speckit.setup (orchestrator)
├── /speckit.setup-constitution
├── /speckit.setup-docs
├── /speckit.setup-skills (new)
├── /speckit.setup-agents
└── /speckit.setup-mcp (new)

Removed intermediate orchestrators (setup-bootstrap, setup-agents as orchestrator).
Each setup-xxx is now a simple sub-command callable individually.

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
Renamed:
- constitution.md → setup-constitution.md
- agents.md → setup-agents.md
- mcp.md → setup-mcp.md

Now all setup commands follow the setup-xxx naming convention:
- /speckit.setup (orchestrator)
- /speckit.setup-constitution
- /speckit.setup-docs
- /speckit.setup-skills
- /speckit.setup-agents
- /speckit.setup-mcp

https://claude.ai/code/session_011KD936jDr39k6crsbjvUty
Add 4-level context hierarchy to reduce functional drift during
implementation and specification:

- Level 0: architecture-registry (unchanged)
- Level 1: specs/__AGENT_CONTEXT_FILE__ (NEW) — project state context
  with domain vocabulary, data model state, interface contracts,
  feature dependency graph, business invariants, cross-cutting concerns
- Level 2: {module}/__AGENT_CONTEXT_FILE__ (ENHANCED) — add interface
  contracts, business invariants, state machines, guard rails,
  module dependency graph sections
- Level 3: {module}/{subdir}/__AGENT_CONTEXT_FILE__ (NEW) — layer-specific
  context for high-complexity directories (threshold-based generation)

Restructure learn.md from 5 to 7 phases, add specs-only/modules-only
scope options, replace hardcoded CLAUDE.md with __AGENT_CONTEXT_FILE__
placeholder for multi-agent support.

https://claude.ai/code/session_01X9jEifqrYMufZ16tWd2zj2
/docs/{domain}/spec.md is the merged, reviewed source of truth
maintained by /speckit.merge. Learn now reads it with higher priority
than specs/ working files:

- Phase 1: new step 4 loads /docs/ before specs/
- Phase 2: each extraction step (vocabulary, entities, contracts,
  invariants, dependencies) prioritizes /docs/ over specs/
- When both define the same term/entity/contract, /docs/ wins
- specs/ supplements with in-progress features not yet merged
- Key Principles: add "Docs over specs" rule
- specs-context-template: reflect dual sourcing in headers and tables

https://claude.ai/code/session_01X9jEifqrYMufZ16tWd2zj2
…n to review

Enhance /speckit.review Phase 2 with 3 new analysis sections:

- 2.2 Dead Code Detection: unused exports, orphaned files, commented-out
  blocks, unused dependencies, stale feature flags. With detection
  scripts and structured report format.

- 2.3 Fake Implementations: TODO/FIXME markers, placeholder returns,
  empty function bodies, mock data in production, stubbed error
  handling, disabled validations, NotImplementedError stubs.
  Classified as Blocker/High/Medium.

- 2.4 Spec Deviation Detection: compares implementation vs spec.md,
  contracts/, data-model.md. Checks requirement coverage (MATCH/DRIFT/
  MISSING/PARTIAL), contract compliance, data model compliance, and
  acceptance criteria verification. Generates compliance score.

Updated: debt categories, report template, action items format with
DRIFT/FAKE/DEAD tags, smart insertion in tasks.md.

https://claude.ai/code/session_01X9jEifqrYMufZ16tWd2zj2
The learn command was ambiguous about when sub-module context files
are generated. Add explicit scope-to-phase activation table:

- Default (no args): does NOT generate sub-modules
- "all": generates sub-modules where threshold is met
- "deep": new scope — forces sub-modules for ALL recognized
  architectural layers (services/, models/, etc.) regardless
  of file count, only 3-file minimum applies
- Feature names / directories: includes sub-modules for those scopes

Add "When to use which scope" guide in Usage Examples.

https://claude.ai/code/session_01X9jEifqrYMufZ16tWd2zj2
Copilot AI review requested due to automatic review settings February 8, 2026 06:33
@Censseo Censseo closed this Feb 8, 2026
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR expands Spec Kit’s template system to provide richer spec/project context artifacts, adds multiple “skills” reference packs, introduces an MCP project-server template, and updates command templates to include semantic anchors and improved workflow guidance.

Changes:

  • Add new spec/plan/idea/feature/context templates and multiple new skill reference bundles.
  • Add MCP “project-server” TypeScript template (process mgmt, API testing, browser automation scaffolding) and update docs to reference new workflows.
  • Update automation scripts (agent-context tech consolidation; bug listing) and adjust packaging/release scripts and Python package include rules.

Reviewed changes

Copilot reviewed 69 out of 69 changed files in this pull request and generated 8 comments.

Show a summary per file
File Description
templates/specs-context-template.md New template for consolidated project state context for specs/ work.
templates/spec-template.md Adds constitution-compliance guidance + new mandatory/conditional sections.
templates/skills/tech-debt/references/refactoring-strategies.md New tech-debt reference content.
templates/skills/tech-debt/references/debt-patterns.md New tech-debt patterns catalog.
templates/skills/tech-debt/SKILL.md New “tech-debt” skill definition + triggers.
templates/skills/microservices-patterns/references/resilience-patterns.md New microservices resilience reference content.
templates/skills/microservices-patterns/references/data-patterns.md New microservices data patterns reference content.
templates/skills/microservices-patterns/SKILL.md New “microservices-patterns” skill definition + triggers.
templates/skills/code-review/references/security-review.md New security code review checklist reference.
templates/skills/code-review/references/review-checklist.md New comprehensive review checklist reference.
templates/skills/code-review/SKILL.md New “code-review” skill definition + triggers.
templates/skills/architecture-patterns/references/implementations.md New architecture implementations reference examples.
templates/skills/architecture-patterns/references/ddd-patterns.md New DDD patterns reference content.
templates/skills/architecture-patterns/SKILL.md New “architecture-patterns” skill definition + triggers.
templates/skills/architecture-decision-records/references/templates.md New ADR templates reference.
templates/skills/architecture-decision-records/references/examples.md New ADR examples reference.
templates/skills/architecture-decision-records/SKILL.md New “architecture-decision-records” skill definition + triggers.
templates/plan-template.md Adds progress tracking checklist section.
templates/module-claude-template.md New module guidelines template (generated by /speckit.learn).
templates/mcp/project-server/tsconfig.json New TS config for MCP server template.
templates/mcp/project-server/src/tools/process.ts New process/log/health/docker tools for MCP server template.
templates/mcp/project-server/src/tools/api.ts New API client/testing utilities for MCP server template.
templates/mcp/project-server/src/config.ts New MCP server config template with placeholders/examples.
templates/mcp/project-server/package.json New MCP server package template (deps/scripts).
templates/mcp/project-server/README.md Documentation for using the MCP server template.
templates/idea-template.md New structured idea template (vision → scope → risks → tech hints).
templates/feature-template.md New decomposed feature template (scope/use-cases/deps/tech hints).
templates/commands/tasks.md Strengthens tasks generation: idea alignment + reuse traceability.
templates/commands/setup.md Adds setup orchestrator command template.
templates/commands/setup-skills.md Adds stack detection → skill generation command template.
templates/commands/setup-mcp.md Adds MCP server generation command template.
templates/commands/setup-docs.md Adds docs initialization command template.
templates/commands/setup-constitution.md Updates constitution setup guidance + propagation expectations.
templates/commands/semantic-anchors.md Adds a central semantic anchors reference.
templates/commands/merge.md Adds merge+docs consolidation command template.
templates/commands/implement.md Adds semantic anchors + agent delegation + reduced upfront context load.
templates/commands/clarify.md Adds semantic anchors + explicit elicitation framing.
templates/commands/checklist.md Adds generate/review modes + constitution/registry auto-check scaffolding.
templates/commands/analyze.md Adds semantic anchors + clearer analysis framing.
scripts/powershell/update-agent-context.ps1 Consolidates/deduplicates tech stacks in agent context updates.
scripts/powershell/list-bugs.ps1 New script to list unresolved validation bugs (PS).
scripts/bash/update-agent-context.sh Consolidates/deduplicates tech stacks in agent context updates (bash).
scripts/bash/list-bugs.sh New script to list unresolved validation bugs (bash).
pyproject.toml Bumps version and ensures templates/scripts/memory are packaged.
memory/constitution.md Reworks constitution template into explicit spec/dev/business sections.
memory/architecture-registry.md Adds architecture registry template for cross-feature consistency.
docs/toc.yml Adds workflows doc to TOC.
docs/quickstart.md Adds guidance for /speckit.change and links to workflow guide.
docs/index.md Links to workflow guide.
README.md Updates command taxonomy + adds “full vs quick” workflow overview.
.github/workflows/scripts/create-release-packages.sh Removes unsupported agents from packaging matrix.
.github/workflows/scripts/create-release-packages.ps1 Removes unsupported agents from packaging matrix (PS).
.github/workflows/scripts/create-github-release.sh Removes unsupported agent artifacts from release creation.
Comments suppressed due to low confidence (1)

templates/commands/setup-constitution.md:50

  • This instruction refers to RATIFICATION_DATE / LAST_AMENDED_DATE, but the updated memory/constitution.md template no longer contains those placeholders (it uses [DATE] for both Ratified/Last Amended). Either update the constitution template to use explicit [RATIFICATION_DATE]/[LAST_AMENDED_DATE] placeholders, or update this command text so it matches the actual template fields.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment on lines +163 to +173
open) ((open++)) ;;
in_progress) ((in_progress++)) ;;
resolved) ((resolved++)) ;;
wont_fix) ((wont_fix++)) ;;
esac

case "$severity" in
critical) ((critical++)) ;;
high) ((high++)) ;;
medium) ((medium++)) ;;
low) ((low++)) ;;
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Because this script uses set -e, the arithmetic increments like ((total++)) / ((open++)) can cause the script to exit early: in bash, ((expr)) returns exit status 1 when the expression evaluates to 0 (e.g., the first total++ evaluates to 0). Use pre-increment (((++total)), ((++open)), etc.) or a form that always returns 0 to avoid unexpected termination.

Suggested change
open) ((open++)) ;;
in_progress) ((in_progress++)) ;;
resolved) ((resolved++)) ;;
wont_fix) ((wont_fix++)) ;;
esac
case "$severity" in
critical) ((critical++)) ;;
high) ((high++)) ;;
medium) ((medium++)) ;;
low) ((low++)) ;;
open) ((++open)) ;;
in_progress) ((++in_progress)) ;;
resolved) ((++resolved)) ;;
wont_fix) ((++wont_fix)) ;;
esac
case "$severity" in
critical) ((++critical)) ;;
high) ((++high)) ;;
medium) ((++medium)) ;;
low) ((++low)) ;;

Copilot uses AI. Check for mistakes.
Comment on lines +176 to +180
let filteredLogs = logs;
if (filter) {
const regex = new RegExp(filter, "i");
filteredLogs = logs.filter((l) => regex.test(l));
}
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

new RegExp(filter, "i") will throw if the user passes an invalid regex (e.g., [), which would crash the tool call. Consider guarding with try/catch and returning a friendly error, or treating the filter as a literal substring (or escaping it) unless the API explicitly promises regex semantics.

Copilot uses AI. Check for mistakes.
Comment on lines +9 to +12
CONSTITUTION COMPLIANCE (v1.1.0)
================================
This spec MUST address the following per .specify/memory/constitution.md:

Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This template references .specify/memory/constitution.md, but other templates in this PR (and the repo’s memory/constitution.md) reference /memory/constitution.md. Please standardize the path to the constitution so generated specs don’t point at a non-existent location depending on install layout.

Copilot uses AI. Check for mistakes.
Comment on lines 135 to 137
- If file exists, append to existing file
- Number items sequentially starting from CHK001
- Each `/speckit.checklist` run creates a NEW file (never overwrites existing checklists)
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There’s an internal contradiction here: step says “If file exists, append to existing file”, but the next bullet says each /speckit.checklist run creates a NEW file (never overwrites existing checklists). Please make the behavior unambiguous (either always create a new uniquely named file per run, or explicitly append/update an existing domain checklist).

Suggested change
- If file exists, append to existing file
- Number items sequentially starting from CHK001
- Each `/speckit.checklist` run creates a NEW file (never overwrites existing checklists)
- If file exists, append to the existing file instead of overwriting it
- Number items sequentially starting from CHK001
- Each `/speckit.checklist` run REUSES the domain file if it exists, appending new items and never deleting or overwriting existing checklist content

Copilot uses AI. Check for mistakes.
Comment on lines +328 to +334
```bash
# Start Docker if configured
docker compose up -d

# Verify services can be started (dry run)
node .mcp/project-server/dist/index.js --test
```
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The validation step suggests running node .mcp/project-server/dist/index.js --test, but the generated MCP server entrypoint doesn’t appear to implement a --test flag. This will lead to a confusing failure during setup. Either add --test handling to the template server, or update the docs to use an existing validation path (e.g., npm run inspect or a documented tool invocation).

Copilot uses AI. Check for mistakes.
}

try {
const frame = page.frameLocator(selector);
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Unused variable frame.

Copilot uses AI. Check for mistakes.
return False

# Determine argument format and output directory based on agent
arg_format = "$ARGUMENTS"
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This assignment to 'arg_format' is unnecessary as it is redefined before this value is used.

Suggested change
arg_format = "$ARGUMENTS"

Copilot uses AI. Check for mistakes.

# Determine argument format and output directory based on agent
arg_format = "$ARGUMENTS"
file_ext = "md"
Copy link

Copilot AI Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This assignment to 'file_ext' is unnecessary as it is redefined before this value is used.

Suggested change
file_ext = "md"

Copilot uses AI. Check for mistakes.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants