profile-bootstrap
Skill0 starsChanged 3 months agoMITView on GitHub
Use the first time this framework is pointed at a project that has no .agentic/profile.md. Reads the repo and drafts a flag taxonomy for PM approval at G1. Without this, a new project runs on a generic taxonomy, which is the same as no taxonomy.
In this repository14 other files it gives its agents.
- capability-scanSkill
- conveneSkill
- decision-recordSkill
- deliverySkill
- escalateSkill
- flag-checkSkill
- haltSkill
- interface-contractSkill
- merge-guardSkill
- roster-preflightSkill
- setback-checkSkill
- setback-logSkill
- stability-checkpointSkill
- work-orderSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 44
If you cannot make a trigger concrete, you have not read enough.
- Very specific: line 62
Write the approved result to .agentic/profile.md and record it as a Decision Record.
≈ “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
Profile bootstrap
The protocol layer of this framework is project-agnostic. The flag taxonomy is not, and it is load-bearing: it decides what gets three models arguing and what gets one model working. A generic taxonomy ("flag risky things") collapses into flagging everything or nothing, and both failures are invisible in the output.
This skill produces a project's taxonomy by reading the project, then asking.
1. Read the repo
In roughly this order, stopping when you have enough:
CLAUDE.md/AGENTS.md/GEMINI.md— many repos already state their own gates, invariants, and forbidden patterns. This is the highest-value file and often contains the taxonomy nearly verbatim.- The setback or incident log, if one exists. A project's recorded failures are the single best predictor of what it will get wrong next. Repeated incident categories are taxonomy entries.
- Docs and design specs — product principles, invariants, non-negotiables. Phrases like "must never", "always requires", "cannot be overridden" mark decisions that are expensive to get wrong.
- Test layout — what the project bothers to test heavily is what it has been burned by.
- Recent git history — commits that revert, fix, or re-fix the same area.
2. Draft the taxonomy
Produce 5-8 triggers. Each must name a concrete category of decision in this codebase, not a virtue.
| Weak | Strong |
| "Flag security-sensitive changes" | "Flag any change to the policy gate's inputs: action type, target, exposure, scope, reversibility" |
| "Flag risky data changes" | "Flag any change to what crosses the school/work/personal boundary" |
| "Flag important architecture" | "Flag any SQL change — it must be checked against both backends" |
If you cannot make a trigger concrete, you have not read enough. Go back to step 1 rather than shipping a vague one.
3. Also record
- Repo path and the branch work happens on
- Acceptance-criteria source — does the PM supply them, or does the Scrum Master derive them? If derived, they must be written into Decision Records.
- Project-specific skills the profile adds under
.agentic/skills/ - Known hazards — anything the environment does that will bite agents (sync-backed directories, restricted networks, dual backends, CRLF sources)
4. Stop at G1
Present the drafted taxonomy to the PM and stop. Nothing runs on a taxonomy the PM has not seen. Their edits are the point — you are proposing, not deciding, and a taxonomy the PM has corrected once is usually right for the project's life.
Write the approved result to .agentic/profile.md and record it as a
Decision Record.
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.