bulma-reviewer
Skill1 starsChanged 51 days agoMITView on GitHub
Graph: review a PR/branch via bulma-code-review, classify CI/test failures, optionally bus a REVIEW_RESULT. Use when /reviewer, /bulma-reviewer, $bulma-reviewer, or a REVIEW_REQUEST envelope arrives.
In this repository35 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/bulma-feature-delivery/adapters/claude.mdCLAUDE.md
- architectSkill
- before-and-afterSkill
- blast-radiusSkill
- bulma-busSkill
- bulma-code-reviewSkill
- bulma-developerSkill
- bulma-fdSkill
- bulma-feature-deliverySkill
- bulma-goalSkill
- bulma-hostSkill
- bulma-lstmSkill
- bulma-memorySkill
- bulma-qaSkill
- bulmaSkill
- bulma-triageSkill
- diagnoseSkill
- figure-it-outSkill
- grill-with-docsSkill
- howSkill
- interrogateSkill
- kevinSkill
- no-commentsSkill
- principle-fix-root-causesSkill
- principle-prove-it-worksSkill
- principle-sequence-verifiable-unitsSkill
- receiving-code-reviewSkill
- tddSkill
- technical-writingSkill
- thermo-nuclear-code-quality-reviewSkill
- to-issuesSkill
- to-prdSkill
- typescript-best-practicesSkill
- unslopSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 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.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 40
Every verdict body includes Model and Reasoning effort rows identifying the executing reviewer's actual settings, or unknown per …
- Very specific: line 51
Do not defer solely because CI could not be read.
- Distinctive: line 55
After posting the verdict, remove this job's clean review-only checkout via JOBS.md review cleanup.
- Very specific: line 59
Does not spawn developer/qa unless the user asks after the report — then write REVIEW_REQUEST is inbound only.
- Very specific: line 63
Not the fd node bulma-fd-reviewer.
≈ “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
Bulma Reviewer (graph)
Orchestrator. Never merge. Chat: bulma-memory. Apply Kevin for human-facing communication.
REQUIRED: GRAPH.md · STATE.schema.md ·
bus PROTOCOL.md · skill bulma-code-review · bulma-memory ·
DISK.md (.bulma-* is global, never git root)
Init .bulma-reviewer/STATE.md. Walk GRAPH (admit first).
code_review is the engine: spec-first, patch bind, high-risk critic.
Its preflight reads all review
context before pass selection and again before publication, pins the reviewed
head, and verifies CI execution. A sensitive-path list stays in the private
job report and does not defer. An unresolved thread whose author approved
this head is not an open concern, and neither is a dismissed change request.
Another reviewer's open change request still gets a code review, then a DEFER
review_pending when the diff is clean. Never
dismiss another review or approve over a concern from someone who did not
approve this head. Every review POST runs behind the engine's scripts/gate.py;
an APPROVE it rejects is never sent. The engine dismisses its own approval
only when a new head gets REQUEST_CHANGES for a code defect. Verdicts are
APPROVE, REQUEST_CHANGES or DEFER; never a COMMENT review, even when another
bot approved. CI that cannot be read is a DEFER ci_unknown (CI cell
not verified (tooling)) unless the diff has a defect.
Review preserves the user's checkout. Local checks use isolated checkouts;
regression claims distinguish source inspection from executed tests. Changes
that depend on another deployment include counterpart and rollout checks via
the engine's evidence rules.
Every verdict body includes Model and Reasoning effort rows identifying
the executing reviewer's actual settings, or unknown per unavailable field.
Second review while main is busy → worktree + general-purpose
worker per PR. --force if CI red. Pending required CI allows pinned review, then waits with retained analysis
(no PR comment) via wait_ci. CI is collected from GitHub Actions
(gh run list -R <repo> --commit <head> and gh run view <id> -R <repo> --json jobs,
plus --log-failed on a failed run) and cross-checked with
statusCheckRollup { state } on the head commit. Never use gh pr checks
or the check-runs / check-suites REST endpoints. If the Actions list and the
rollup query both fail, including HTTP 403, continue the review and mark CI
unverified.
Do not defer solely because CI could not be read. Draft / conflict / CI-red DEFER is
expected, and the draft gate is unchanged.
Cursor and Claude invoke /bulma-reviewer. Codex invokes $bulma-reviewer.
Worktree and branch rules: JOBS.md §Worktree.
After posting the verdict, remove this job's clean review-only checkout via
JOBS.md review cleanup. Pending CI or
failed publication keeps it for resume.
Does not spawn developer/qa unless the user asks after the report —
then write REVIEW_REQUEST is inbound only; outbound fix = tell the
user or write PR_BUG_FIX only if authorized and the issue is the PR.
Not the fd node bulma-fd-reviewer.
Verified external concerns feed FEEDBACK.md: attribute actual stage gaps, save scoped prevention, and draft only actionable improvement tasks. A request for changes or LSTM invocation alone proves no miss. Issue publication requires authorized policy and an exact destination.
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.

