agentleFS
UnfoldCRO-OrganizationUnfoldCRO-Organization/cro-agent-skills/skills/project-start/SKILL.md

project-start

Skill0 starsChanged 22 days agoMITView on GitHub

Start, resume or switch to a named project using minimum required context when the user says work on, resume, continue, open, switch to or check a project.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

27 rules. 0 are common, 6 are distinctive, and 21 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 21

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If yes, do not read the global project index.

    line 18
  2. Very specific:

    If no, read only the matching entry from ~/.claude/project-index.md.

    line 19
  3. Very specific:

    Use ~/.claude/context.md only as fallback when the project cannot be resolved from the lightweight index.

    line 20
  4. Very specific:

    Always read the project index: <project>/CLAUDE.md, or <project>/.claude/CLAUDE.md where the docs are local-only (git repo with a remote).

    line 28
  5. Very specific:

    Never go looking for a slot marked unwritten.

    line 34
  6. Very specific:

    If there is no Docs map (legacy project), read selectively and offer setup-project-context migrate once:

    line 36
  7. Very specific:

    Read <project>/rule.md only when permanent unique project rules affect the task.

    line 37
  8. Very specific:

    Read <project>/docs/task.md only when continuing current work or checking active implementation.

    line 38
  9. Very specific:

    Read <project>/docs/pending.md only when selecting next work or reviewing blockers.

    line 39
  10. Very specific:

    Read <project>/docs/scope.md only when validating whether work is included.

    line 40
  11. Very specific:

    Read <project>/docs/proposal.md only when checking commercial commitments or original timelines.

    line 41
  12. Very specific:

    Read <project>/docs/completed-vs-scope.md only for delivery status, handover or scope-completion checks.

    line 42
  13. Very specific:

    Read <project>/docs/changelog.md only for confirmed change history or release reporting.

    line 43
  14. Very specific:

    Read <project>/docs/deploy.md only for push, release, deployment, rollback, tagging or production work.

    line 44
  15. Very specific:

    Read <project>/context.md only when prior technical state is required.

    line 45
  16. Very specific:

    Read <project>/DESIGN.md only for UI, layout, component or responsive work.

    line 46
  17. Very specific:

    Read <project>/memory.md only for recurring issues or known gotchas.

    line 47
  18. Very specific:

    Read <project>/spec.md only for architecture, API, schema, authentication, permissions, calculations, security or core workflows.

    line 48
  19. Very specific:

    Check .claude/rules/ only for rules matching the affected work area.

    line 50
  20. Very specific:

    Check .claude/skills/ only when a unique project workflow is required.

    line 52
  21. Distinctive:

    Do not invoke implementation capabilities during project start unless the same request clearly requires specialised work.

    line 57
  22. Very specific:

    When capability selection is required after project resolution, use task-dispatch.

    line 58
  23. Distinctive:

    Do not read files only because they exist.

    line 62
  24. Distinctive:

    Do not load unrelated project folders.

    line 63
  25. Distinctive:

    Do not load full registry information when the requested project is already identified.

    line 64
  26. Very specific:

    Do not read spec.md for ordinary copy, styling or layout work.

    line 65
  27. Distinctive:

    Do not read infrastructure or deployment files unless those actions are requested.

    line 66

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Start or Resume Project Work

Requested project or task: $ARGUMENTS

Objective

Resolve the requested project and load only the documentation required to begin the current task.

Process

  1. Determine whether the current working directory is already the requested project.
    • If yes, do not read the global project index.
    • If no, read only the matching entry from ~/.claude/project-index.md.
    • Use ~/.claude/context.md only as fallback when the project cannot be resolved from the lightweight index.
  2. Resolve only necessary information:
    • Project name.
    • Local project path.
    • Platform or short stack summary.
    • Environment, hosting or deployment information only when the task requires it.
  3. Always read the project index: <project>/CLAUDE.md, or <project>/.claude/CLAUDE.md where the docs are local-only (git repo with a remote). If it has a ## Docs map (v2 project), read only the mapped slots the task needs, plus the slot's .local.md where the map says +local:

    • resume / next work → docs/08-IMPLEMENTATION.md (+ .local.md §Checkpoint)
    • UI, layout, components → 02 · structure, refactor → 03 · schema → 04 · endpoints, webhooks → 05
    • agent workflow, project skills → 06 · URLs, access, client → 07 · local setup → 09 · feature work → 10 (+ 01 for scope)
    • auth, permissions, security → 11 · MCP tooling → 12 · third-party services → 13 · tests, QA → 14
    • bug, recurring issue → 15 · push, release, deploy, rollback → 16 · change history → 17
    • Never go looking for a slot marked unwritten; create it from the code only when the task needs it (standard: ~/.claude/references/project-docs-standard.md).

    If there is no Docs map (legacy project), read selectively and offer setup-project-context migrate once:

    • Read <project>/rule.md only when permanent unique project rules affect the task.
    • Read <project>/docs/task.md only when continuing current work or checking active implementation.
    • Read <project>/docs/pending.md only when selecting next work or reviewing blockers.
    • Read <project>/docs/scope.md only when validating whether work is included.
    • Read <project>/docs/proposal.md only when checking commercial commitments or original timelines.
    • Read <project>/docs/completed-vs-scope.md only for delivery status, handover or scope-completion checks.
    • Read <project>/docs/changelog.md only for confirmed change history or release reporting.
    • Read <project>/docs/deploy.md only for push, release, deployment, rollback, tagging or production work.
    • Read <project>/context.md only when prior technical state is required.
    • Read <project>/DESIGN.md only for UI, layout, component or responsive work.
    • Read <project>/memory.md only for recurring issues or known gotchas.
    • Read <project>/spec.md only for architecture, API, schema, authentication, permissions, calculations, security or core workflows.
  4. Check .claude/rules/ only for rules matching the affected work area.
  5. Check .claude/skills/ only when a unique project workflow is required.

Capability Rule

  • Project resolution does not require a specialist agent, MCP or plugin.
  • Do not invoke implementation capabilities during project start unless the same request clearly requires specialised work.
  • When capability selection is required after project resolution, use task-dispatch.

Token Rules

  • Do not read files only because they exist.
  • Do not load unrelated project folders.
  • Do not load full registry information when the requested project is already identified.
  • Do not read spec.md for ordinary copy, styling or layout work.
  • Do not read infrastructure or deployment files unless those actions are requested.

Initial Response After Resolution

State only:

Project resolved: <project-name>
Docs: v2 | legacy
Context loaded: <files actually read>
Task focus: <requested task or next confirmed task>

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.