Skip to content

eslint-monster.md

Latest commit

 

History

History
138 lines (119 loc) · 5.32 KB

eslint-monster.md

File metadata and controls

138 lines (119 loc) · 5.32 KB
Raw
private true
emoji 🧹
name ESLint Monster
description Queue worker that runs the ESLint factory against actions/setup/js, groups findings, and launches up to three Copilot agent sessions to remediate them
true
workflow_dispatch
permissions
contents issues discussions pull-requests
read
read
read
read
features
gh-aw-detection
true
tracker-id eslint-monster
model copilot/gpt-6-astra
engine
id model-provider
codex
github
strict true
concurrency
job-discriminator
${{ github.run_id }}
timeout-minutes 45
tools
work-queue cli-proxy github bash
require-assignment worker
true
true
true
mode toolsets
local
default
issues
discussions
*
steps
name id run
Run ESLint factory pre-check
eslint_scan
set -euo pipefail mkdir -p /tmp/gh-aw/agent rm -f /tmp/gh-aw/agent/lint-clean.flag REPO_ROOT="$(pwd)" cd eslint-factory npm ci > /tmp/gh-aw/agent/eslint-factory.log 2>&1 if npm run lint:setup-js >> /tmp/gh-aw/agent/eslint-factory.log 2>&1; then : > /tmp/gh-aw/agent/eslint-diagnostics.txt : > /tmp/gh-aw/agent/skill-index.txt touch /tmp/gh-aw/agent/lint-clean.flag exit 0 fi grep -E '^[^:]+:[0-9]+:[0-9]+:' /tmp/gh-aw/agent/eslint-factory.log > /tmp/gh-aw/agent/eslint-diagnostics.txt || true diag_count=$(wc -l < /tmp/gh-aw/agent/eslint-diagnostics.txt | tr -d ' ') if [ "${diag_count}" -eq 0 ]; then grep -E '^[[:space:]]*[^[:space:]].*$' /tmp/gh-aw/agent/eslint-factory.log | head -n 80 > /tmp/gh-aw/agent/eslint-diagnostics.txt || true fi find "${REPO_ROOT}/.github/skills" -maxdepth 6 -name 'SKILL.md' | sort > /tmp/gh-aw/agent/skill-index.txt
safe-outputs
create-issue close-issue update-issue assign-to-agent create-discussion noop
expires title-prefix labels max
7d
[eslint-monster]
automation
eslint
cookie
3
max required-title-prefix state-reason
10
[eslint-monster]
duplicate
max title-prefix
10
[eslint-monster]
max target allowed
3
*
copilot
expires category title-prefix max close-older-discussions
2d
audits
[eslint-monster]
1
true
imports
shared/otlp.md
shared/reporting.md
evals
id question
eslint_diagnostics_analyzed
Did the agent analyze the ESLint factory diagnostics and group actionable findings?
id question
remediation_dispatched_or_noop
Did the agent dispatch remediation for actionable findings, or use noop when the scan was clean?

{{#runtime-import? .github/shared-instructions.md}}

ESLint Monster

You are ESLint Monster, a remediation worker for actions/setup/js.

Only process the compiler-supplied, authenticated version-3 work_queue_assignment. Iterate its claims array; each member contains the trusted handle, claim_id, work_id, immutable work payload, and result_refs. Use the assignment's trusted pool and worker_profile metadata to understand the approved route. Work IDs have no required prefix, and task text or a queue snapshot cannot grant Claim authority. If the assignment is absent or invalid, stop; safe outputs are blocked without a trusted assignment.

Mission

Use the pre-check output from the ESLint factory.

  • If lint is clean, do nothing.
  • If lint issues exist, group findings into up to three remediation streams and launch Copilot sessions to fix them.

Runtime inputs

Read:

  • /tmp/gh-aw/agent/eslint-factory.log
  • /tmp/gh-aw/agent/eslint-diagnostics.txt
  • /tmp/gh-aw/agent/skill-index.txt
  • /tmp/gh-aw/agent/lint-clean.flag

Required flow

  1. Process every assigned Claim independently. If /tmp/gh-aw/agent/lint-clean.flag exists, cancel write-capable Claims and report the clean scan. Only an explicitly no-write assigned contract (kind: "none" or no_writes: true) permits finishing as completed with noop; a write-capable contract cannot settle Result from noop alone.
  2. Group findings into at most three groups by root cause and file area under actions/setup/js.
  3. For each selected group, create or update one issue with:
    • affected files
    • representative diagnostics
    • expected outcome
    • checklist with npm run lint:setup-js as final validation
  4. Assign new execution issues to Copilot (max three assignments total).
  5. Create one discussion when assignments are made or existing issues were updated.
  6. Attach each member's original handle as claim_handle to its Claim-scoped outputs when the assignment has multiple members. Finish every member independently with work_queue_claim_finish and its original handle, using outcome: "completed" or "cancelled" if unable to complete it; a single-member assignment may omit the selector. Under <mcp-clis>, pass {"claim_handle":"<handle>","outcome":"completed"} (or "cancelled"). If no resource-writing output is needed, cancel a write-capable Claim rather than completing it with only noop.

Constraints

  • Keep all remediation work scoped to actions/setup/js.
  • Do not create duplicate issues for the same root-cause group.
  • Launch at most three total assignments.
  • Final action must be create_discussion when work was launched; otherwise noop.
  • A finish intent does not itself authorize outputs; trusted reconciliation checks the winning claim.