contribute

name	contribute
description	Generalize a lesson and stage it for Memnyx contribution. Delegates sanitization to `sanitizer` before staging.
user_invocable	true
args	Optional description of the lesson or improvement to generalize

Model Selection

See .claude/skills/_shared/MODEL_SELECTION.md (in your workspace) for full policy.

• Default model: Sonnet 4.6 — generalizing a project-specific fix into a reusable pattern needs judgment about what's portable
• Promote to Sonnet when: never — generalization is the core task
• Promote to Opus when: never

Contribute — Generalize Lessons for Memnyx

You are preparing a generalized contribution that can be pulled into Memnyx. The contribution must be project-agnostic — no project names, file paths, domain terms, API keys, or internal details.

Setup — Resolve <workspace> and scope: The skill's base directory is <workspace>/.claude/skills/contribute/; walk up three directory levels and validate that <workspace>/.claude/.workspace exists. Abort with a setup-broken error if validation fails. Then read the active session marker from <workspace>/sessions/active/*.md and <workspace>/projects/*/sessions/active/*.md to determine scope: if project_slug is workspace, scope is <workspace>; otherwise scope is <workspace>/projects/<project_slug>. If no marker exists, ask the user whether to read from workspace or a specific project before proceeding.

Steps

1. Gather input: If a lesson or improvement was passed as argument, use it. Otherwise, read <scope>/.claude/memory/lessons-learned.md and ask the user which lessons they'd like to contribute back.

2. Classify the contribution into one of these types:

   • skill-update — improvement to an existing skill's instructions
   • new-skill — a new skill that would benefit all projects
   • convention — a general coding or workflow convention
   • memory-template — improvement to memory file structure or initial content
   • docs — improvement to documentation in .claude/docs/
   • config — improvement to .mcp.json, settings.json, or CLAUDE.md

3. Generalize: Rewrite the lesson to remove all project-specific details:

   • Replace project names with [project] or generic terms
   • Replace specific file paths with generic patterns (e.g., src/auth/login.ts → "authentication modules")
   • Replace domain-specific terms with generic equivalents
   • Replace specific tool/framework names only if the lesson applies broadly; keep them if the lesson is framework-specific
   • Keep the actionable insight intact — the generalized version must be just as useful

4. Sanitize (MANDATORY): Write the draft to a temp path and invoke sanitizer on it:

   /sanitizer <temp-draft-path> --mode=boilerplate
   

   If sanitizer returns any findings (SECRET, PII, PRIVATE_CONTEXT, or TONE), do not proceed. Surface the full report to the user and apply recommended fixes before continuing. The sanitizer is the dedicated scrubber — do not duplicate its logic here.

5. Write the contribution: Create a new file in <workspace>/contributions/ with this format:

   Filename: YYYY-MM-DD-<short-slug>.md (e.g., 2025-06-15-improve-hello-mcp-retry.md)

   Content:

   # Contribution: <title>
   
   **Type**: skill-update | new-skill | convention | memory-template | docs | config
   **Target**: <which file(s) in Memnyx this would affect>
   **Date**: YYYY-MM-DD
   
   ## Problem
   
   <What problem or gap was discovered>
   
   ## Recommendation
   
   <The generalized improvement, with specific text/changes to apply>
   
   ## Rationale
   
   <Why this improves Memnyx, based on real project experience>
   

6. Review with user: Show the contribution along with the sanitizer report and ask:

   • "Sanitizer pass — findings above. Any other adjustments needed?"
   • "Ready to save, or want to adjust?"

7. Save: Write the file to <workspace>/contributions/.

8. Report:

   Contribution staged
   ===================
   File: <workspace>/contributions/<filename>.md
   Type: <type>
   Target: <target files>
   
   To integrate into Memnyx, run /pull-contributions from the Memnyx repo
   and point it at <workspace>/contributions/.
   

Important Rules

• Never include: project names, internal URLs, API endpoints, team names, proprietary domain terms, specific file paths from the project
• Always include: the actionable recommendation, the category, which Memnyx files would change
• Ask before saving: Always show the user the generalized version before writing it