Github ops excellence
Skill chf3198/megingjord-harness/skills/github-ops-excellence
AI agent governance harness: baton workflow, fleet LLM routing (Ollama/Claude/OpenRouter), and CI gates for Copilot, Claude Code, and Codex.
npx -y skills add chf3198/megingjord-harness --skill github-ops-excellenceAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Define and apply expert GitHub policy profiles and control catalogs used by specialized execution skills across ticketing, review/merge, governance, and release flows.
SKILL.md
6.8 KB, as published. Nobody here has run it
GitHub Ops Excellence
Purpose
Provide a bounded, auditable policy and control catalog for high-quality GitHub execution across repositories.
Hard constraints
- No unbounded loops or recursive retries.
- Maximum one full pass per invocation.
- Maximum ten actionable recommendations per invocation.
- No silent policy changes; propose changes with rationale.
- No bypass escalation without explicit approval trail.
- If required evidence is missing, return
NO_CHANGEwith missing artifacts.
Modes
triage: issue quality and planning readiness.refinement: backlog readiness and dependency hygiene.pre-pr: branch hygiene and PR preparation.pre-merge: review/check/ruleset merge readiness.sprint-health: flow health, WIP risk, and review throughput.release-readiness: release gate and rollback readiness.incident-flow: containment-to-follow-up issue flow during production incidents.admin-audit: governance/ruleset/admin posture review.
Scope boundary (catalog ownership)
This skill is the policy catalog owner, not the primary executor.
Primary execution belongs to specialist skills selected by github-ops-tree-router.
This skill provides:
- policy-profile calibration (
strict|standard|light) - shared control definitions
- severity and verification standards
- conflict resolution guidance when specialists disagree
This skill is not the primary owner for repository profile/discoverability hygiene
(topics, homepage, social preview, community health file completeness). Delegate those to repo-profile-governance.
Invocation order
- For plan/feature-sensitive controls, run
github-capability-resolverfirst. - Run
github-ops-tree-routerto select execution skill(s). - Apply
github-ops-excellenceas profile overlay to calibrate strictness and evidence requirements.
Shared control packs by mode
triage
- Issue title is a plain imperative sentence ≤72 chars — no
type(scope):prefix (Conventional Commits belongs on commits/PRs, not issues). - Issue title has no pseudo-prefixes (
[BUG],[P1],TICKET-123,JIRA-123) and does not duplicate type in the title when atype:*label exists. - GitHub
#Nis used as canonical ID; no parallel local ID schemes (TICKET-NNN, bracket tags, etc.) present. - Issue includes: problem, expected outcome, acceptance criteria.
- Metadata includes: assignee, labels (priority + area + type), milestone/iteration, project link.
- Large work is decomposed with sub-issues/dependencies.
- Template/form adherence is verified;
blank_issues_enabled: falseconfirmed in repo.
refinement
- Backlog item has definition-of-ready (scope, acceptance, dependencies, risk).
- Parent/child issue relationships are coherent (sub-issues, dependencies).
- Priority and iteration fields are populated.
- Blocked work has explicit owner and next action.
pre-pr
- One branch per concern.
- PR links issue(s) and states test evidence.
- PR template fields completed.
- Reviewers and code owners requested where applicable.
pre-merge
- Required reviews satisfied.
- Required status checks passing on latest commit.
- Conversation resolution complete.
- Ruleset/branch-protection requirements satisfied.
- If merge queue is required, CI supports
merge_grouptrigger.
sprint-health
- WIP is within agreed team limits.
- Aged items are identified with owner and resolution plan.
- PR review latency and merge latency are measured against team targets.
- Blocked-item ratio is tracked and triaged.
release-readiness
- Changelog/release notes prepared.
- Release evidence links present (tests/checks/artifacts).
- Rollback or remediation path documented.
- No unresolved blocking issues for target milestone.
incident-flow
- Incident item includes severity, impact, owner, and containment plan.
- Hotfix branch/PR is linked to incident issue and validation evidence.
- Rollback trigger and rollback owner are explicit.
- Follow-up issues (root cause, prevention, docs/tests) are created before closure.
admin-audit
- Rulesets/protection target critical branches.
- CODEOWNERS coverage exists for protected paths (deep ownership hygiene belongs to
repo-profile-governance). - Bypass list is minimal and justified.
- Ruleset history/insights reviewed for drift.
- Security gates (code scanning/dependency review) enforced where required.
actions-security
GITHUB_TOKENleast privilege baseline is enforced.- Third-party actions pinning policy is enforced.
- OIDC is preferred for cloud auth over long-lived static secrets.
- Workflow ownership is protected (
.github/workflows/under CODEOWNERS). - Runner trust boundaries are documented and policy-compliant.
Output format (required)
GITHUB_OPS_POLICY_REPORT
mode: <triage|refinement|pre-pr|pre-merge|sprint-health|release-readiness|incident-flow|admin-audit>
scope: <repo|team|org|enterprise>
policy_profile: <strict|standard|light>
controls:
- id: F1
severity: <low|medium|high>
area: <ticketing|branching|pr|governance|security|release|projects|actions-security>
control: <policy control statement>
strict: <requirement>
standard: <requirement>
light: <requirement>
calibration_actions:
1) priority: <P1|P2|P3>
executor_skill: <selected specialist>
profile_adjustment: <specific calibration>
verification: <objective pass condition>
metrics_snapshot:
- metric: <lead-time|cycle-time|review-latency|merge-latency|blocked-age|wip|reopen-rate>
observed: <value>
target: <value>
status: <ok|breach>
decision:
- <apply|defer|NO_CHANGE>
missing_evidence:
- <none or required artifacts>
Evidence requirements
Collect only artifacts relevant to mode and selected specialist path:
- Issues/PR metadata
- Branch and ruleset/protection settings
- Check run results and review states
- Project field/status snapshots
- Project insights/charts or equivalent flow snapshot
- Release notes and milestone status
Stop conditions
Return NO_CHANGE when:
- Evidence is incomplete for the selected mode.
- Recommendation cannot be objectively verified.
- Proposed change would reduce required safety controls.
- Same recommendation was recently applied and verified.
- Metrics are unavailable and no reliable proxy exists.
Quality bar
A valid result must be:
- Specific (not generic)
- Testable (clear pass/fail)
- Minimal (least disruptive change)
- Traceable (links finding to action)