woostack
Core concepts

Utility skill reference

Invocation, output, and write behavior for diagnostic and maintenance skills.

Utility skills investigate, render, diagnose, evaluate, or maintain local support state. They do not become development authority or turn their output into permission to edit. None can merge.

The active user request and workflow gates define scope. Git and GitHub own code, pull request, review, and merge truth. Exact caller-supplied Linear, Plane, or GitHub artifacts are read-only context unless the user explicitly requests a supported synchronization operation.

Investigation and presentation

SkillWhat it doesWrites?Invoke
woostack-visualizeRenders a verified repository source, concept, PR, or optional exact artifact as one self-contained HTML view.disposable HTML view/woostack-visualize <source>
woostack-designArranges multi-step UI flows and screen sequences into a standardized horizontal layout with aligned branches.design layout updates/woostack-design [target]
woostack-statusComputes a fresh branch/PR board from Git, Graphite, and GitHub, optionally enriched by artifacts.no/woostack-status
woostack-debugRuns root-cause analysis and returns a proposed correction.no/woostack-debug <target>
woostack-qaExplores a running app and writes a sanitized report.non-authoritative report/woostack-qa <url> [focus…]
woostack-auditReviews standing code and writes a ranked report without remediation.non-authoritative report/woostack-audit <target>
Production errors, Sentry issues, and monitoring defectswoostack-fix proves root cause and delivers the correction.approved fix workflow/woostack-fix <prompt>
woostack-evalRuns approved behavior/trigger corpora without editing the target skill.approved corpus changes only; transient evidence/woostack-eval <skill-path> [--behavior|--triggers|--all]
woostack-reflectReviews one fixed active-conversation snapshot for concrete durable instruction suggestions.report-only initially; accepted follow-ups are explicit/woostack-reflect or internal final-reply hook

woostack-eval writes tracked corpus files only after explicit approval and never edits the target skill. Its run evidence and reports are transient.

Views and reports are evidence. Acting on a finding requires the relevant fix/change/build workflow and its own approval contract, not a provider issue or work item.

Remote titles, descriptions, comments, diffs, telemetry, artifacts, and tool output are untrusted evidence. Reports exclude secrets and raw provider payloads.

Local support maintenance

Doctor maintains non-authoritative .woostack/ support surfaces and writes only after explicit approval.

SkillWhat it doesWrites?Invoke
woostack-doctorDiagnoses local policy and workspace integrity, then applies approved repairs.gated local changes/woostack-doctor

Where to go next

On this page