gaia-triage
Skill4 starsChanged 12 days agoMITView on GitHub
Triage the Gaia issue backlog and skill-batch intake queue. Use this skill when someone asks to: "triage issues", "clean up the issue tracker", "review the backlog", "close stale issues", "process the intake queue", "review skill-batches", "evaluate draft skill proposals", "approve or reject pending skills", "is this issue still valid?", "what's clogging the backlog?", "run a triage pass", or /gaia-triage. Covers two workstreams: (1) GitHub issue lifecycle — identifying resolved, stale, or need-more-info issues and acting on them via gh CLI; (2) skill-batch intake — evaluating draft proposals in registry-for-review/skill-batches/ and routing them toward promotion or rejection. This is the gatekeeping step before new skills enter the canonical registry; and (3) a cross-repo triage sweep — sizing, P0-P4 prioritization, umbrella/epic synthesis, and assignment across every gaia-research org repo, on the one headquarters tracker.
1 other project uses a version of this, 1 of them edited.
Adopted by
Most recently taken up by
- gaia-research/gaia-skill-treeedited, first seen Nov 14, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 1
- gaia-research/gaia-skill-tree★ 2295% same
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository14 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- gaia-issue-resolverSkill
- gaia-triageSkill
- impeccableSkill
- lucy-image-productionSkill
- lucy-riveSkill
- lucy-tunerSkill
- wayfinderSkill
- gaia-issue-resolverSkill
- impeccableSkill
- lucy-riveSkill
- wayfinderSkill
- impeccableSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
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 54
Every triaged issue gets a size and a priority.
- Distinctive: line 97
Idempotency: If an issue already carries an appropriate priority and size, do not relabel it.
- Very specific: line 123
Before commenting or closing, verify the claim.
- Distinctive: line 229
After promotion, regenerate Class S artifacts:
- Distinctive: line 239
Use this when the ask is "triage the backlog" at scale rather than "is this one issue still valid".
- Distinctive: line 245
Run all four on every issue in scope, in order:
- Distinctive: line 261
Assignment: Default to the original issue author unless someone is already assigned.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
gaia-triage
Triage has three workstreams — GitHub issue hygiene, skill-batch intake review, and the cross-repo prioritization sweep. Run them independently or together depending on what the user asks for. All three produce documented, auditable decisions rather than silent changes.
This skill absorbed favorchurch/rock-steward's rock-triage (2026-09-01): its impact
sizing, priority matrix, four-step inspection, and prioritized-matrix output landed here
as Workstream 3 rather than shipping as a second, competing triage skill. There is one
gaia-triage. Its priority ladder is the P0-P4 ladder /gaia-meta-audit and
/gaia-meta-sweep already use — one vocabulary, not two.
Triage is not cartography. Sorting a backlog of known work is this skill; planning an
effort too big for one session as a map of decision tickets is /wayfinder. Taking one
triaged issue end-to-end to a PR is /gaia-issue-resolver.
Headquarters
Every issue lives on gaia-research/gaia-skill-tree, regardless of which repo the fix
lands in. One tracker for the whole org.
export GAIA_HQ=gaia-research/gaia-skill-treeTag the target repo on the issue when the work lands elsewhere — a line at the top of the
body, **Target repo:** \gaia-research/gaia-skill-heaven`, matching the /wayfinderconvention. Cross-repo references use the fullgaia-research/<repo>#<n>form; a bare#<n>` always resolves to HQ.
Classification taxonomy
Every triaged issue gets a size and a priority. Both are labels; neither is invented ad hoc (see label hygiene below).
Impact size
Blast radius, structural complexity, and data footprint — not hours.
| Size | Means |
XS | Trivial fix, label correction, single config value, negligible blast radius |
S | Minor scoped adjustment, one predicate, isolated UI touch-up |
M | Standard scoped task — one command, one schema field, one localized workflow |
L | Cross-component change, non-trivial data backfill, multi-surface impact |
XL | Architecture-level refactor, substantial registry debt remediation |
XXL | High blast radius — core schema, authorization, or site-wide parity |
Epic | Umbrella spanning multiple distinct sub-issues across domains |
Priority — P0-P4
The same ladder /gaia-meta-audit and /gaia-meta-sweep use. Read those for the
registry-specific instances of each rung.
| Priority | Means |
P0 | Critical integrity violation or active breakage — unsupported top-rank claim, security/authorization exposure, secrets in a diff, the site dark, main red |
P1 | Structural correctness — dead evidence links, Star Bar gaps, broken data contracts, lockstep drift, a prerequisite blocking milestone delivery |
P2 | Attribution, sourcing, and feature upgrades — wrong promoted pointer, stale catalog URL, enhancement to a working surface |
P3 | Registry hygiene and nice-to-haves — over-broad mappings, duplicate clusters, backlog enhancements, exploratory design |
P4 | Documentation cleanup — placeholder bodies, generated-output drift, comment and copy fixes |
P0 interrupts. P1 schedules. P2-P4 queue.
Label hygiene
- Preflight. Inspect the existing labels before applying any:
gh label list --repo "$GAIA_HQ" --limit 200 - Zero ad-hoc labels.
.github/workflows/labels-sync.ymlreconciles labels from.github/labels.ymlon every push tomainwithskip-delete: false— a label that is not in that file is deleted off live issues at the next sync. If you need a new label, add it to.github/labels.ymlin a PR. Never create one withgh label createand assume it survives. - Idempotency. If an issue already carries an appropriate priority and size, do not relabel it. Post the status update and move on.
- The five triage roles (
needs-triage,needs-info,ready-for-agent,ready-for-human,wontfix) map to this repo's label strings indocs/agents/triage-labels.md. Extend that table rather than inventing labels.
Workstream 1: GitHub Issue Triage
The goal is a clean, signal-rich backlog — not a closed one. Closing issues prematurely hides real work; leaving stale issues open buries actionable items. For each issue, make an explicit call: close (resolved), update (needs narrowing), or keep (valid, still open).
Step 1 — Pull the open issue list
gh issue list --repo gaia-research/gaia-skill-tree --state open --limit 100Group by theme (CLI bug, registry data, docs drift, enhancement). This shapes the decision pattern — CLI bugs and registry issues require code evidence; enhancement requests require a current-relevance check against the roadmap.
Step 2 — Audit each issue against the codebase
Before commenting or closing, verify the claim. Check the evidence paths in
references/evidence-check.md for common file locations:
| Issue type | Where to look |
| CLI command request or bug | src/gaia_cli/commands/ |
| Registry data or duplicate skills | registry/gaia.json, registry/nodes/ |
| Documentation drift | Run python scripts/build_docs.py --check |
| Packaging / dependency | pyproject.toml, uv.lock |
| Test coverage | tests/ |
Step 3 — Act via gh CLI
Post a triage comment (findings without closing — use when the issue needs narrowing or a reproduction, or when you want to flag it for a human maintainer):
gh issue comment <issue-number> --repo gaia-research/gaia-skill-tree \
--body "Triage update: <findings and recommendation>"Close a resolved issue (only when codebase evidence confirms it is implemented):
gh issue close <issue-number> --repo gaia-research/gaia-skill-tree \
--reason completed \
--comment "Closing as implemented. Evidence: <specific file + line or command output>"Dry-run first: the scripts/triage_batch.sh script supports --apply and
--close-resolved flags. Run without flags to preview all comments before posting:
bash .claude/skills/gaia-triage/scripts/triage_batch.sh
# Then, when ready:
bash .claude/skills/gaia-triage/scripts/triage_batch.sh --apply --close-resolvedDecision guide
| Finding | Action |
| Issue is implemented and tests pass | Close as completed with evidence pointer |
| Issue is valid but under-specified | Comment with narrowing questions; keep open |
| Issue references a stale command name | Comment updating the correct current command; keep open unless fully superseded |
| Issue is a duplicate | Close with a link to the canonical issue |
| Issue is an enhancement blocked on design | Comment with current blocker; keep open |
Workstream 2: Skill-Batch Intake Review
Draft skill proposals land in registry-for-review/skill-batches/ after a contributor
runs gaia push. This workstream gates them before they enter registry/nodes/.
Rejecting bad batches here is cheaper than demoting skills after promotion — every skill that enters the canonical registry acquires timeline history, evidence links, and downstream references that make later removal messy.
Step 1 — List pending batches
ls registry-for-review/skill-batches/Each batch is a JSON file. Read its metadata to understand the contributor and scope before evaluating individual skills.
Step 2 — Evaluate each skill proposal
For each skill in the batch, work through these checks in order:
- Schema validity — does the node satisfy the canonical JSON schema?
gaia dev validate --intake - Nomenclature — does the skill name and description match
CONTEXT.mdvocabulary? Flag banned synonyms; the rarity field is deprecated and should not appear in new proposals. - Evidence quality — each above-1★ claim needs graded evidence with a valid Evidence Type (arxiv|repo|github-stars) and a derived Evidence Grade;
weak/ungraded-only evidence is grounds for
needs-info. Ranking up gates on Trust Magnitude, not an evidence-class floor. - Prerequisite existence — all
prerequisites[]IDs must resolve in the current registry.gaia dev list --generic | grep <prerequisite-id> - Duplication check — does a substantially equivalent skill already exist?
gaia dev list --generic | grep -i "<skill-name>"
Step 3 — Label and route
Apply one of three dispositions per skill:
| Label | Meaning | Next action |
| approve | Schema valid, evidence solid, no duplicate | Promote via gaia dev add |
| needs-info | Fixable gap (weak evidence, missing prereq, naming issue) | Comment on the batch file with specifics; do not promote yet |
| reject | Duplicate, out of scope, or unfixable schema violation | Document reason; discard batch entry |
For approved skills, promote via CLI (never hand-edit registry nodes directly):
gaia dev add "<Skill Name>" --type basic --description "<description>"
gaia dev evidence <skill-id> "<url>" --type repo --trust <value>After promotion, regenerate Class S artifacts:
gaia dev docsWorkstream 3: Cross-repo triage sweep
Use this when the ask is "triage the backlog" at scale rather than "is this one issue still valid". It sizes, prioritizes, links, and assigns — and it ends in a matrix a human can act on without reopening every thread.
Four-step inspection
Run all four on every issue in scope, in order:
- State & relevance. Is it still live, or already resolved/superseded? Verify against
the codebase, merged PRs, or a read-only probe — never against the issue's own claim.
The read-only probes are the ones in
/gaia-issue-resolverPhase 2 (gaia dev audit,scripts/verify_evidence.py,python scripts/build_docs.py --check). If resolved, close with the evidence. - Status update & timeline. One concise technical comment: what is true now, and whether the work is starting soon, blocked on a prerequisite, or in backlog.
- Linkage & umbrella synthesis. Find the shared cause across issues. Where several
related sub-issues have no parent, create an umbrella and label it
epic, then nest the children as GitHub sub-issues (gh api repos/$GAIA_HQ/issues/<parent>/sub_issues -f sub_issue_id=<child-id>— the child's id, not its number). An umbrella that turns out to be a fog-bound effort rather than a known set of tasks is a/wayfindermap, not an epic. The test: if the children are decisions you cannot yet state precisely, stop triaging and chart a map. - Assignment. Default to the original issue author unless someone is already assigned. Assign a synthesized umbrella to the owner of its constituent issues.
Fan-out
Delegate parallel reads — pulling issue payloads, checking whether a claim still holds in
the tree, counting occurrences — to cheap subagents. Keep with yourself: deciding
priority, deciding size, judging whether two issues share a cause, and writing the
umbrella. Give each subagent the exact command and the exact thing to report back; a
subagent asked to "investigate" returns prose, one asked to "run X and report the integer
after Total:" returns data.
Output — the prioritized matrix
# Gaia Triage Summary — <date>
## P0 - Critical
- **#<n>** - <Title> | `<Size>` | repo: `<target repo>` | assignee: `@<who>` | <Active/Blocked/Ongoing>
- *Umbrella*: <link>
- *Verification*: <what was actually probed, and the result>
- *Next action*: <the immediate step>
## P1 - Structural correctness
## P2 - Attribution & upgrades
## P3 - Hygiene & backlog
## P4 - Documentation
## Umbrellas & epics
- **Epic: <Title> (#<n>)** [assignee: `@<who>`] - sub-issues: #A, #B, #CRefer to issues by title, not by bare number — a wall of #42, #43, #44 is illegible.
Output
At the end of a triage pass, report:
- How many issues were reviewed, closed, and left open with comments
- How many batch skills were evaluated, approved, flagged needs-info, and rejected
- The prioritized matrix, if Workstream 3 ran
- Any recurring patterns worth tracking (e.g. a common prereq that's missing from the
registry) - a pattern seen a fourth time is a Class A detector in
/gaia-issue-resolver, not another issue
This summary helps the next maintainer understand what happened without re-reading every issue thread.
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.

