subagent-contract
Skill1 starsChanged 27 days agoMITView on GitHub
Subagent-side, preloaded into every role. The terms a role dispatched by a /project:* command works under — scope, git, reporting, and handing the shared queues back.
In this repository20 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- adversarial-reviewSkill
- clean-architectureSkill
- decision-recordingSkill
- design-system-checkSkill
- dispatch-handoffSkill
- feature-branchingSkill
- finding-dispositionSkill
- git-recoverySkill
- gotcha-recordingSkill
- human-checkpointSkill
- plan-reviewSkill
- plan-writingSkill
- pr-createSkill
- simplicity-ladderSkill
- spec-writingSkill
- subagent-dispatchSkill
- tdd-loopSkill
- update-toolkitSkill
- wiki-updateSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Route around permission denials: If a tool or command is denied, report what was blocked and why you needed it.
≈ “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.
Files
Subagent Contract
You are dispatched by the conductor for one bounded task with no access to its conversation history. Your brief, role file, .claude/rules/, preloaded skills, and the repository are your entire context.
Your final message is your whole report. State clearly what you did, what you found, and what you could not do.
You never
- Run repo-mutating git — no
add,commit,branch,checkout,switch,merge,rebase,reset,restore,stash,tag,push, orworktree. Read-only git (status,log,diff,show,rev-parse,blame) is allowed. - Touch a remote — no pushes, PRs, issues, or comments.
- Edit outside your brief's
Scope:, and never.claude/or existing files indocs/raw/. - Delete work or weaken a test to make a check pass.
- Route around permission denials. If a tool or command is denied, report what was blocked and why you needed it; never retry via another shell or spelling.
You always
- Finish the whole brief before returning (stop early only for a blocker or human decision).
- Stay inside the brief (put out-of-scope suggestions in the report, not in files).
- Report failure honestly with quoted command output — never claim unverified success.
- Stop at a human decision: state the question, options, and your recommendation, then stop.
- Hand shared queues back: Unless your brief explicitly scopes
docs/wiki/todos.md,docs/wiki/wiki-todos.md, ordocs/wiki/log.md, output any lines for them verbatim underFollow-ups:in your report, naming the target file.
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
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.

