agentleFS
tenforward-audiotenforward-audio/swe-harness/.agents/skills/develop-project/SKILL.md

develop-project

Skill0 starsChanged 3 months agoMITView on GitHub

Implement, fix, refactor, test, or change dependencies after the user explicitly authorises an unrecorded mutation or a bounded In progress lane. Use for code, configuration, behavior-linked documentation, dependency evaluation, and isolated implementation. Do not use to advance a recorded item from intake or Planning, manage lifecycle state, investigate, technically review, integrate, or release.

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 15 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈15

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the root AGENTS.md, contributing guide, and only the other policies relevant to the change.

    line 20
  2. Distinctive:

    Test important success, invalid-input, operational-failure, preservation, and supported-compatibility paths at the owning boundary.

    line 25
  3. Distinctive:

    Run focused tests and the affected-layer gate while iterating, then the configured full check-and-build gate once for a green checkpoint.

    line 29
  4. Distinctive:

    Do not execute unresolved command placeholders.

    line 32
  5. Distinctive:

    Follow the compatibility boundary in root instructions.

    line 37
  6. Distinctive:

    When project instructions mark prototype state disposable, update all current producers, consumers, tests, and fixtures together and …

    line 41
  7. Distinctive:

    Do not expand lane scope, mutate another lane's files or stateful resources, or edit canonical queues and status boards from a background …

    line 52
  8. Distinctive:

    Follow the style guide for naming, modules, errors, comments, and tests.

    line 61
  9. Distinctive:

    Keep framework glue, domain behavior, and infrastructure boundaries distinct.

    line 63
  10. Distinctive:

    Prefer explicit data and state over speculative abstractions or duplicate sources of truth.

    line 64
  11. Distinctive:

    Read the dependency policy before adding, replacing, copying, or bundling external code or assets.

    line 66
  12. Distinctive:

    Read the plugin capability policy before making an external plugin a required or approved project capability.

    line 68
  13. Distinctive:

    Read the security policy for sensitive boundaries, permissions, private data, untrusted input, or supply-chain changes.

    line 70
  14. Very specific:

    Create or supersede an ADR under ../../decisions/ only for a consequential, durable choice with credible alternatives and reversal …

    line 72
  15. Distinctive:

    Update public documentation, decisions, dependency records, security guidance, and the Unreleased changelog when their claims change.

    line 75
  16. Distinctive:

    Leave the implementation branch pinned to that commit.

    line 82

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Develop the Project

Execution mode

inline

For a recorded issue or feature, require an In progress parent card or an assigned implementation lane. Route a request to work on, start, or implement an item still in intake or Planning through deliver-project-work. Do not independently move cards or edit canonical queues and boards.

Implement from the owning boundary

  1. Read the root AGENTS.md, contributing guide, and only the other policies relevant to the change.
  2. Inspect current implementation, behavior, and tests before selecting an approach. Trace input, state, failures, and output through the owning layer.
  3. Implement the smallest complete behavior while preserving unrelated work.
  4. Test important success, invalid-input, operational-failure, preservation, and supported-compatibility paths at the owning boundary.
  5. For interactive or user-visible behavior, exercise it when available and report observed and unobserved evidence separately from automated checks.
  6. Run focused tests and the affected-layer gate while iterating, then the configured full check-and-build gate once for a green checkpoint.

Do not execute unresolved command placeholders. Ask for the missing command only when verification depends on it.

Respect actual compatibility obligations

  • Follow the compatibility boundary in root instructions. Do not preserve a historical shape without a named supported release, current consumer, authoritative data set, staged rollout, or explicit constraint. Record the scope, failure behavior, and removal condition of a temporary obligation.
  • When project instructions mark prototype state disposable, update all current producers, consumers, tests, and fixtures together and rebuild or reset it.
  • Preserve current-state integrity: validation, atomic updates, crash recovery, cache invalidation, and version identifiers are not speculative legacy support.

Work inside assigned isolation

  • A mutating lane created by coordinate-parallel-work must use its recorded common base, worktree, branch, dependencies, and owned boundaries.
  • Do not expand lane scope, mutate another lane's files or stateful resources, or edit canonical queues and status boards from a background worktree.
  • Return checkpoint commits, verification evidence, conflicts, and material review notes to the coordinator.
  • Parallel permission does not authorise merge, branch deletion, history rewriting, push, or remote changes.

Apply engineering governance

  • Follow the style guide for naming, modules, errors, comments, and tests.
  • Keep framework glue, domain behavior, and infrastructure boundaries distinct. Prefer explicit data and state over speculative abstractions or duplicate sources of truth.
  • Read the dependency policy before adding, replacing, copying, or bundling external code or assets.
  • Read the plugin capability policy before making an external plugin a required or approved project capability.
  • Read the security policy for sensitive boundaries, permissions, private data, untrusted input, or supply-chain changes.
  • Create or supersede an ADR under ../../decisions/ only for a consequential, durable choice with credible alternatives and reversal conditions.

Update public documentation, decisions, dependency records, security guidance, and the Unreleased changelog when their claims change. Finish with the verification, isolation, staged-diff review, and checkpoint procedure in the contributing guide.

Return the exact candidate commit, checks, and observed or unobserved evidence to deliver-project-work or manage-project-work for the Reviewing transition. Leave the implementation branch pinned to that commit; lifecycle tracking is a separate checkpoint on the canonical coordinating branch.

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.