Skip to content

feat: overhaul agent suite, hooks, and CLI types#3

Merged
melagiri merged 7 commits intomasterfrom
feature/agents-hooks-overhaul
Feb 14, 2026
Merged

feat: overhaul agent suite, hooks, and CLI types#3
melagiri merged 7 commits intomasterfrom
feature/agents-hooks-overhaul

Conversation

@melagiri
Copy link
Copy Markdown
Owner

Summary

  • Agents: Enhance technical-architect with LLD standards, ADR templates, schema verification, expert pushback table. Add 4 new agents: fullstack-engineer (replaces cli-engineer + web-engineer), ux-designer, product-manager (Jira-free, uses GitHub Issues), journey-chronicler (open-source framing)
  • Hooks: Update branch-discipline and parallel-warning to reference new agent names. Add no-jira block hook to prevent Jira API calls
  • Types: Add customTitle?: string to ParsedSession, 'prompt_quality' to InsightType union
  • CLAUDE.md: Update agent suite table, document ownership, hookify rules, and repo structure
  • Plans: Include design and implementation plan documents

Test plan

  • Verify all 5 agents exist in .claude/agents/ with correct YAML frontmatter
  • Verify old agents (cli-engineer.md, web-engineer.md) are removed
  • Verify 6 hooks exist in .claude/ including new hookify.no-jira.local.md
  • Verify cli/src/types.ts has customTitle?: string on ParsedSession and 'prompt_quality' on InsightType
  • Verify CLAUDE.md agent suite table matches actual agents
  • Run cd cli && pnpm build to confirm no type errors

🤖 Generated with Claude Code

melagiri and others added 7 commits February 14, 2026 09:42
…back

Add new sections to the Technical Architect agent definition:
- LLD Standards: 500-line max, modular directory structure, content rules table
- Schema Alignment Verification: pre-approval checklist, red flags table, output format
- Architecture Decision Records (ADR) template and guidance
- Firestore Performance Patterns: read/write patterns, anti-patterns to reject
- Type Evolution Strategy: adding, extending, deprecating, renaming fields
- Technology Guardrails: locked choices table
- Expanded Expert Pushback: premature scaling, contradictory requirements,
  missing considerations, scope creep, gold plating, bikeshedding
- Enhanced Communication Style: progressive disclosure, anchoring to examples
- Collaboration guidance with other agents
- Expanded Context Sources table

All existing content preserved. Only additive changes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ourney-chronicler

Create four new agent definitions to replace cli-engineer and web-engineer
with a unified agent suite:

- fullstack-engineer (sonnet): Replaces cli-engineer and web-engineer.
  Principal engineer working across both CLI and web repos. Includes
  development ceremony steps, CI gate, cross-repo type alignment rules,
  error handling patterns, debugging workflow, PR template, and testing
  strategy guidance.

- ux-designer (opus, cyan): Code-first UX design without Figma. Includes
  Developer Dev and Team Lead Taylor personas, ASCII wireframe format,
  user flow format, screen specification template, research framework,
  quality checklist, and handoff process to fullstack-engineer.

- product-manager (sonnet, green): Jira-free PM using GitHub Issues and
  local markdown tracking. Includes ceremony coordination, MoSCoW
  prioritization, progress reporting format, risk management, feature
  request triage, communication templates, and scope management.

- journey-chronicler (opus, amber): Captures learning moments, breakthroughs,
  and course corrections. Includes trigger types, entry format, tag taxonomy,
  five thematic arcs, suggest+approve pattern, quality gates, voice
  guidelines, and genericization guide for open-source sharing.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ra hook

- branch-discipline: Expand agent pattern from (cli-engineer|web-engineer)
  to include fullstack-engineer and ux-designer

- agent-parallel-warning: Update sequential patterns to reflect new agent
  names (PM -> Fullstack Engineer, TA -> Fullstack Engineer, PM -> TA)

- no-jira (NEW): Block hook that prevents Jira/Atlassian API calls.
  Matches createJiraIssue, jira_create_issue, atlassian.net, and ENGG-
  patterns. Directs to GitHub Issues and local tracking instead.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove the separate cli-engineer and web-engineer agents, now replaced
by the unified fullstack-engineer agent.

Update CLAUDE.md:
- Agent Suite table: replace cli-engineer and web-engineer with
  fullstack-engineer, ux-designer, product-manager, journey-chronicler
- Document Ownership table: update agent references
- Hookify Rules table: add no-jira block rule
- Repository structure comment: update agent list

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add optional `customTitle` field to ParsedSession interface for
  user-defined session titles (supports rename from web dashboard)
- Add 'prompt_quality' to InsightType union for prompt effectiveness
  analysis feature

Both changes are backward compatible (optional field, union extension).
Web repo types.ts will need matching update for cross-repo alignment.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add the design and implementation plan documents for the agents/hooks
overhaul and analysis features phase:
- Design plan: agent definitions, hook changes, type changes
- Implementation plan: phased execution strategy

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… patterns

- Add prompt_quality to TA agent insight type enforcement rule
- Add prompt_quality to CLAUDE.md insight type references (lines 88, 348)
- Remove stale deleted agent names (cli-engineer, web-engineer) from
  branch-discipline hook pattern

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@melagiri melagiri merged commit d379a7e into master Feb 14, 2026
@melagiri melagiri deleted the feature/agents-hooks-overhaul branch February 14, 2026 11:09
melagiri added a commit that referenced this pull request Mar 9, 2026
…nce to prompts

- CANONICAL_PATTERN_CATEGORIES: 8 canonical kebab-case categories
- EFFECTIVE_PATTERN_CLASSIFICATION_GUIDANCE: per-category disambiguation text
- SESSION_ANALYSIS_SYSTEM_PROMPT: category field added to effective_patterns schema
- FACET_ONLY_SYSTEM_PROMPT: same update for backfill prompt
- AnalysisResponse type: category field required (not optional)
- JSON schema examples updated with concrete structured-planning example
- Observability warning for LLM missing category field (mirrors friction-monitor)

Wording from LLM Expert finalized spec (Task #3).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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.

1 participant