ea-approve
Skill3 starsChanged 3 months agoMITView on GitHub
Promote reviewed model content from staging to approved -- the only write path into the approved zone. Use when a human asks to approve, promote, accept or publish staged model changes, or after a review session concludes staged content is ready. Never use on your own initiative; promotion asserts human sign-off.
Files
Promoting staging to approved
model/approved/ is human-signed content: downstream governance, reporting and
context extracts read only from it. Promotion is the act of a person putting their
name on a proposal. Your job is to make that act safe and informed -- never to
perform it unprompted.
The one rule that is not negotiable
Do not promote unless the human has explicitly asked for this promotion, now. "The validator passes" is not an instruction to promote. If you finished modelling work and staging is clean, report that it is ready and stop. When in doubt whether an instruction covers promotion, ask.
Installed as a plugin, the host enforces both halves of this. This skill cannot be
loaded on a model's own initiative -- the human invokes it (/ea-skills:ea-approve) --
and a hook refuses any write into model/approved/ that does not go through easkills
promote. If the hook refuses one of your writes, you were about to bypass the gate:
put the change in model/staging/ and stop there.
Procedure
- Dry-run the gate:
easkills promote --root <repo> --dry-runThe gate validates approved + staging as if already merged, by approved-zone standards: ownership and review dates become errors, every semantic rule is on. This is stricter than a plain staging validation -- content that warns in staging blocks here, by design.
- Present what would change. List the files that would move, the concepts they
add or update (a same-path file replaces the approved one -- that is the update
flow), and every open
PROV006assumption riding along. Assumptions do not block promotion, but promoting one means the human accepts it -- say so explicitly rather than letting it slide through in a file list. - On explicit confirmation, promote:
easkills promote --root <repo> # everything
easkills promote --root <repo> --file model/staging/x.yaml # selected filesPartial promotion is normal: promote what was reviewed, leave the rest staged.
- Commit immediately. The move is a plain file rename; the git commit is the approval record (who, when, what). Suggest a message naming what was approved and on whose decision. Do not batch unrelated changes into it.
When the gate blocks
The usual causes, in order of frequency:
GOV001/GOV002-- staged content missingownerorlastReviewed. These were warnings in staging and are errors now. Get the real owner from the human; never invent an address to make the gate pass.PROV*-- evidence problems that were tolerated during drafting. Fix them in staging, re-run the dry-run.- Semantic errors surfacing only in the merged picture (a staged relationship whose approved endpoint was retyped, a duplicate name against approved content).
Fix in staging and re-run. Never edit model/approved/ directly to resolve a
promotion conflict -- if approved content is wrong, that correction is itself a
staged proposal with its own review.
Reporting back
State what was promoted (files and concepts), what was deliberately left in staging and why, which assumptions the human accepted by promoting, and the commit that records it. If you were asked to promote and refused, say exactly which rule blocked it and what evidence is missing.
In this repository25 other files it gives its agents.
- ea-adrSkill
- ea-ai-governanceSkill
- ea-alignSkill
- ea-boardSkill
- ea-capability-mapSkill
- ea-change-triageSkill
- ea-checkSkill
- ea-complianceSkill
- ea-contextSkill
- ea-delta-ingestSkill
- ea-dispensationSkill
- ea-docsSkill
- ea-evalSkill
- ea-healthSkill
- ea-importSkill
- ea-intakeSkill
- ea-modelSkill
- ea-regulatorySkill
- ea-runSkill
- ea-securitySkill
- ea-serviceSkill
- ea-stakeholdersSkill
- ea-standards-baseSkill
- ea-validateSkill
- ea-viewsSkill
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.

