About ziw-setup
ziw-setup creates or refreshes workflow configuration for people maintaining repository workflows. It verifies mappings against repository and tracker evidence, updates adapters, and reports gaps affecting specific workflows.
Key features
- 01Create or refresh workflow config
- 02Verify fields within a defined scope
- 03Query tracker metadata
- 04Update workflow adapters
- 05Report gaps and verification results
Use cases
- 01Set up workflow config for a repository
- 02Verify config after workflow changes
- 03Update adapter references to config
MCP support
The skill calls for using a configured MCP complaint writer where applicable; it does not describe other MCP connectivity.
Install
Copy the command and run it in a terminal.
npx skills add https:/ / github.com/ zaks-io/ skills --skill ziw-setupAfter installing, type/ziw-setupin Claude Code to confirm it loaded.
Bundled files
References and scripts installed along with SKILL.md.
agents/openai.yamlreferences/agent-workflow.mdhandoff.mdissue-tracker-contract.mdlinear-cursor-example.mdoperating-profile.mdproject-config.mdDetails
From the project READMESetup
Create or refresh docs/agents/workflow/config.md, the workflow lookup table.
Finish with verified mappings, working adapters, and gaps blocking only affected workflows.
Inputs
Repo path, existing config/rules, and requested setup or refresh scope.
Steps
- Read existing config. Initially use project-config.md as the field index; load applicable sections. On refresh, patch existing config.
- Set verification scope from the request and drift evidence.
- Verify values through repo files and read-only provider queries.
- Write stable config, preserving conventions and decisions.
- Update adapters; report checks, changes, and gaps.
Set verification scope
- Run initially, then refresh for workflow changes or requested verification.
- Initial setup or explicit full verification covers every applicable field.
- Scoped refresh covers changed/suspect fields and dependencies. Preserve unrelated stable mappings without claiming revalidation; optional unrelated integrations do not block.
- Stale commands, renamed labels, moved projects, CI, delegation, or environment changes expand only affected scope.
- For in-scope tracker mappings, make a read-only query with configured IDs/query-safe names. Replace stale names only with verified equivalents for the same scope.
- Change missing, stale, wrong, or newly verified fields. Preserve explicit human decisions unless the user or current evidence contradicts them.
Load references only for matching work:
- Tracker states, labels, readiness, bodies, estimates: issue-tracker-contract.md.
- Workers, unattended hooks, continuation, loop limits, capacity, merge safety: operating-profile.md.
- Linear + Cursor: linear-cursor-example.md.
- Roles/adapters: agent-workflow.md.
- Handoffs: handoff.md.
- Unmapped glossaries: glossary-discovery.md.
Friction: file encountered friction with the configured complaint tool, once per distinct issue and without secrets. Load friction-log.md when the writer is missing, fails, or replies ambiguously; for tracker-based intake; or for orchestrator event categories or a run rollup.
Verify values
Use current evidence within scope:
- Repo rules, package scripts/lockfiles, Make/Just/Turbo, Git, and CI establish identity, commands, generated artifacts, branch/PR conventions, and gates. Query code-host protections, linked checks, and PR query shapes when needed.
- Specs, glossaries/context maps, ADRs, runbooks, and user instructions establish planning authority, paths, status conventions, and documentation checks.
- Read-only tracker calls establish location, projects/boards/teams, statuses, labels, priorities, estimates, relations, templates, and query contracts. Record exact query-safe names, stable IDs, and accepted fields.
- Tracker metadata, verified config, or user instructions establish delegation, environment labels, continuation handles, and remote mechanics. Record supported paths rather than a full assignee inventory.
- Resolve target-repo Claude/Codex/editor files, imports, symlinks, and generated metadata from a clean path. Verify distribution source/mode, lockfile, refresh command, project paths, symlink layout, and marketplace.
- Deployment config, CI, runbooks, service inventories, and user instructions establish preview/hosted-check rules, credentials, and production approval.
- Verify enabled hosted review from provider config/docs/settings.
CodeRabbit root
.coderabbit.yamlmapsreviews.auto_review.
Verify commands from definitions when sufficient; report unexecuted commands. Read-only metadata queries use existing authorized access. Omit unverifiable values and name missing evidence. Track authorized actionable follow-ups; otherwise report owner/action. Required gaps block only affected workflows.
Gate parity
- Map hooks, local checks, and required CI jobs to commands. Flag missing local tasks; preserve thresholds, environment requirements, and secret-scan scope.
- Trust configured task caches; list only specific tasks whose keys miss inputs.
- Verify unattended enforcement and production approval through the operating profile.
- Prefer the configured MCP complaint writer. Inspect its schema without fictitious events; tracker fallback needs explicit configuration.
Write stable config
- Keep reusable values, paths, commands, IDs, routing, limits, and policies terse. Include applicable planning, tracker/readiness, delegation, access, review/merge, environment safety, complaint, and handoff mappings.
- Tracker taxonomy supplies defaults, not evidence labels exist. Preserve verified conventions and align separate label docs with config.
- Keep readiness, risk, review evidence, type, area, and ownership policy consistent.
- Preserve glossary paths. New defaults are
GLOSSARY.mdand, across contexts,GLOSSARY-MAP.md. - Record project, plugin/marketplace, managed, user/global, or mixed skill distribution.
- Classify generated
ziw-*copies as committed dependencies, symlink fanout, ignored cache, absent, or repo-authored. Retain source/lockfile for committed generated dependencies and update mechanically. - Store progress, blockers, baseline failures, and verification follow-ups in the tracker. Read CI/PR/review/deploy health and provider availability live; names/IDs describe schema.
- Before removing state notes, search prior setup titles/descriptions with a bounded, tracker-location/repo-route query. Transfer nonduplicates to configured intake without readiness labels when authorized. Otherwise preserve notes and report the action.
Update adapters
- Point existing
AGENTS.md,CLAUDE.md, editor rules, and usage docs to config before workflow skills. Use the role reference to keep adapters short. - Follow the target repo's Claude registration rules and import its agent Markdown,
usually one-line
@AGENTS.md. - Symlink only required exact paths; verify targets from a clean checkout. Report refused paths and track repairs without guessing.
Never
- Invent or retain contradicted values, or activate disabled optional integrations.
- Store state snapshots, execution evidence, tasks, transcripts, or
Unknownsin config. - Discard state awaiting transfer or substitute a local backlog without tracker-write authority.
- Assign real implementation issues to probe workers or file fictitious complaints.
- Hand-edit generated downstream skills.
- Include secrets, tokens, signed URLs, customer payloads, or private logs in config.
- Run install, deploy, expensive hosted, or credentialed provider actions solely for setup verification without explicit approval. Authorized read-only metadata queries are allowed.
- Deploy or mutate production during setup.
Done
Report config path, initial/full/scoped mode, fields verified/changed/preserved, adapters, commands, executed checks, gaps, authorized follow-ups, and affected-workflow readiness. Scoped refresh does not establish whole-config verification.