Merged
Conversation
- MCP client test: write temp server script inside project root so Bun can resolve node_modules, and add env/cwd passthrough to StdioClientTransport - Go skill runner test: increase timeout from 5s to 30s for CI environments where go compilation is slower https://claude.ai/code/session_01P4uwywxub339LH15Y2wKVL
Ignore temp directory created by MCP client test. https://claude.ai/code/session_01P4uwywxub339LH15Y2wKVL
The Go test timeout was a symptom of missing node_modules, not actual compilation overhead. go run completes in under 1s locally. https://claude.ai/code/session_01P4uwywxub339LH15Y2wKVL
The Go test was reverted thinking it was a node_modules issue, but go run compiles from scratch on CI (no build cache), which genuinely exceeds the default 5s bun test timeout. Set per-test timeout to 15s. https://claude.ai/code/session_01P4uwywxub339LH15Y2wKVL
chinmaymk
pushed a commit
that referenced
this pull request
Mar 21, 2026
…Prompt - Detailed Edit tool guidance: read first, exact matching, uniqueness, multi-line blocks — addresses the #1 source of editing failures - Edit vs Write decision guidance - Environment discovery instructions: read package.json, check git status, find configs on first interaction (replaces deleted detect-environment script) https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h
chinmaymk
added a commit
that referenced
this pull request
Mar 21, 2026
* feat: add ra-claude-code recipe — coding agent with Claude Code system prompt principles Distills Claude Code's 110+ conditional prompt components into 6 focused skills: - claude-code-agent: core behavior (read-before-write, safety, tool discipline, output style) - context-engineer: project context discovery (CLAUDE.md, AGENTS.md, conventions) - planner: task decomposition for complex multi-step work - debugger: systematic reproduce → isolate → fix → verify workflow - git-workflow: safe git ops (new commits, explicit staging, no force push) - code-style: correctness → security → simplicity → readability Includes environment detection script, token budget middleware, security checklist reference, and auto-compaction at 80% context threshold. https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * fix: convert detect-environment script from bash to TypeScript for Windows compatibility https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * feat: add 6 new skills, context discovery, compaction prompt, and memory to ra-claude-code New skills: - auto-mode: execute immediately, don't over-ask, data exfiltration prevention - verify: type-check → lint → test → build after every change - quick-commit: Claude Code's exact git commit protocol (parallel assess, HEREDOC, stage-by-name) - quick-pr: PR creation with full branch analysis and gh pr create - explore-delegate: when/how to spawn subagents for codebase exploration - stuck-recovery: detect loops, challenge assumptions, force strategy changes New activation script: - context-engineer/scripts/discover-context.ts: walks cwd → git root, finds and injects CLAUDE.md/AGENTS.md/.cursorrules content at session start Config changes: - Custom compaction prompt preserving file paths, decisions, git state, task progress - Memory persistence enabled - All 12 skills registered https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * refactor: move core behavior to systemPrompt, remove redundant skills, tune descriptions - Remove context-engineer skill (ra handles context discovery OOTB) - Remove claude-code-agent and auto-mode skills — merged into systemPrompt so core behavior is always active without needing skill activation - Remove app.skills from config (no longer supported) - All 9 remaining skills are on-demand via skillDirs discovery - Rewrite all skill descriptions as trigger-based ("Use when...") so the model knows exactly when to activate each skill - Update README to reflect new architecture https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * feat: add file editing guardrails and environment discovery to systemPrompt - Detailed Edit tool guidance: read first, exact matching, uniqueness, multi-line blocks — addresses the #1 source of editing failures - Edit vs Write decision guidance - Environment discovery instructions: read package.json, check git status, find configs on first interaction (replaces deleted detect-environment script) https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * docs: rewrite ra-claude-code README for clarity Punchy, scannable format. Leads with what the agent does, not how it's built. Covers editing reliability, environment discovery, and project rule auto-detection. https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * feat: rename planner→plan skill with approval gate, add todo skill - Renamed `planner` to `plan` so users can invoke it as `/plan` - Plan mode now saves to scratchpad and blocks implementation until user explicitly approves - New `todo` skill tracks a checklist in the scratchpad, survives context compaction https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h * Rewrite todo skill to use per-task scratchpad keys Each task gets its own scratchpad entry (todo#1, todo#2, etc.) instead of a single key holding the full list. This makes updates atomic — completing or removing one task no longer requires rewriting the entire list, eliminating the risk of dropping items during read-modify-write. Also adds instruction to always show the list to the user after updates. https://claude.ai/code/session_015S8JmSrsFZk4hYdNgeZF5h --------- Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.