{"id":51118658,"url":"https://github.com/kreek/consult","last_synced_at":"2026-06-25T00:01:12.939Z","repository":{"id":353649171,"uuid":"1220277357","full_name":"kreek/consult","owner":"kreek","description":"Engineering quality focused skills for AI coding agents that keep humans in the loop.","archived":false,"fork":false,"pushed_at":"2026-06-05T15:52:40.000Z","size":31757,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-05T18:15:38.876Z","etag":null,"topics":["agentic","ai-agents","claude-code","claude-code-plugin","claude-skills","codex","codex-plugin","cursor","cursor-plugin","skills"],"latest_commit_sha":null,"homepage":"","language":"TypeScript","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/kreek.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":"SECURITY.md","support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null}},"created_at":"2026-04-24T18:23:24.000Z","updated_at":"2026-06-05T15:52:44.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/kreek/consult","commit_stats":null,"previous_names":["kreek/agent-booster-pack","kreek/has-skills","kreek/consult"],"tags_count":4,"template":false,"template_full_name":null,"purl":"pkg:github/kreek/consult","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kreek%2Fconsult","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kreek%2Fconsult/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kreek%2Fconsult/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kreek%2Fconsult/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/kreek","download_url":"https://codeload.github.com/kreek/consult/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/kreek%2Fconsult/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34753781,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-24T02:00:07.484Z","response_time":106,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agentic","ai-agents","claude-code","claude-code-plugin","claude-skills","codex","codex-plugin","cursor","cursor-plugin","skills"],"created_at":"2026-06-25T00:01:11.045Z","updated_at":"2026-06-25T00:01:12.928Z","avatar_url":"https://github.com/kreek.png","language":"TypeScript","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Consult\n\nDrawn from 25 years of software engineering across startups and large\norganizations, Consult is a human in the loop focused portable skill library for\nraising the engineering maturity of coding agents.\n\nHumans are good at mapping real-world issues to technical solutions, and, given enough context,\ncoding agents are good at generating a correct implementation. Consult does not try to\nchange either side of that equation: it does not automate humans away, and it\ndoes not change how coding agents, or their harnesses, work internally. Instead, it\naugments the collaboration between them so agent-assisted work produces simpler,\ntrustworthy, and maintainable production-grade software while humans still own\nintent, design, and acceptance.\n\nConsult is not a vibe-coding safety net. It works best when you have\nenough software fundamentals to judge plans, tradeoffs, risk, and proof. Consult can\nstill suggest a solution, but it's more effective if you can make the hard\nengineering decisions yourself (plus you'll actually know how the code works).\n\n## What Consult guides agents to do\n\n- Keep humans in the loop for significant and hard-to-change work: public\n  interfaces, project structure, dependency picks, data boundaries, long-lived\n  behavior, substantial new modules, non-trivial logic, and deliberate behavior\n  changes.\n- Use AI to improve the developer's mental model, not to replace it.\n- Model data first: make values, states, and rules clear; limit side effects;\n  keep state changes at the edges.\n- Show their work: prove behavior with tests, contracts, logs, or visible\n  checks rather than relying on what seems correct.\n- Treat security, data safety, and accessibility as essential, not extras.\n- Plan beyond launch: invest in observability, reliability, safe deployment,\n  and a rollback plan.\n- Organize work into clear, reviewable changes humans can trust and maintain.\n\n## Install\n\nMost users should install the package for their primary agent. Use the manual\ninstall only when you want one shared skill directory across several tools, or\nwhen your tool does not support plugins.\n\n### Claude Code\n\nInside Claude Code:\n\n```text\n/plugin marketplace add kreek/consult\n/plugin install consult@consult\n```\n\n### Codex\n\n```sh\ncodex plugin marketplace add kreek/consult\n```\n\nThen open `/plugins` in Codex, find **Consult** in the list,\npress Enter to open its details, and select `Install plugin`.\n\nTo update:\n\n```sh\ncodex plugin marketplace upgrade consult\n```\n\nThen open `/plugins`, find **Consult** in the list, press Enter\nto open its details, and select `Install plugin` again.\n\n### Cursor\n\nInstall from the [Cursor Marketplace](https://cursor.com/marketplace) (search for\n**Consult** or submit this repo at\n[cursor.com/marketplace/publish](https://cursor.com/marketplace/publish) if it is\nnot listed yet). Open the marketplace panel in Cursor, install **consult**, then\nconfirm skills under **Settings → Rules → Agent Decides**. Invoke a skill with\n`/skill-name` in Agent chat (for example `/workflow`, `/proof`).\n\nTo test from a local checkout before marketplace listing:\n\n```sh\nmkdir -p ~/.cursor/plugins/local\ncp -R /path/to/consult/plugin ~/.cursor/plugins/local/consult\n```\n\nReload the window (**Developer: Reload Window**). Prefer `cp -R` over symlinks;\nsome Cursor builds do not load symlinked local plugins reliably.\n\nIf you also run `./setup.sh`, Cursor can load the same skills twice (plugin plus\n`~/.agents/skills/`). Use either the Cursor plugin or manual install for Cursor,\nnot both.\n\nDeveloping inside this repository with a local plugin copy also duplicates skills\n(project `agents/.agents/skills/` plus the plugin). See\n[`CONTRIBUTING.md`](CONTRIBUTING.md#local-plugin-development).\n\n### Pi\n\n```sh\npi install git:github.com/kreek/consult\n```\n\n`github:` is not a Pi package source prefix; use `git:` for GitHub shorthand.\nAfter installing, run `/reload` inside Pi. Consult includes bundled skills plus\nruntime extensions for `/proof` and `/consult:self-review`.\n\n### Google Antigravity\n\nAntigravity's CLI (`agy`) manages plugins with `agy plugin install`. Consult ships a\nready-to-install plugin at `plugin/` (a `plugin.json` marker plus a `skills/`\ndirectory). Install it from a local checkout:\n\n```sh\nagy plugin install /path/to/consult/plugin\n```\n\n`agy` copies the plugin into `~/.gemini/antigravity-cli/plugins/consult` and\nregisters its skills; verify with `agy plugin list`. Because it copies rather\nthan links, re-run the command (or `./setup.sh`) after pulling new skills.\n`./setup.sh` runs this automatically when `agy` is on your PATH.\n\n### Manual (multi-agent or unsupported plugins)\n\nPrerequisites: Git and GNU Stow. Install Stow with one of:\n\n```sh\nbrew install stow   # macOS\napt install stow    # Debian/Ubuntu\ndnf install stow    # Fedora/RHEL\n```\n\n```sh\ngit clone https://github.com/kreek/consult.git\ncd consult\n./setup.sh\n```\n\n`setup.sh` prints the actions it will take and confirms before changing\nanything. It links `~/.agents/skills/`, installs the Antigravity plugin via\n`agy plugin install`, and links tool-specific skill locations when those tools\nare present. End-user installs do not need Python or uv.\n\n## Skills\n\nConsult includes 24 skills. Open a skill for its triggers, workflow, and\nverification.\n\n- Routing and proof:\n  - [`workflow`](agents/.agents/skills/workflow/SKILL.md): Use first for almost every software engineering task to route risks, choose skills, and define proof.\n  - [`proof`](agents/.agents/skills/proof/SKILL.md): Tests, claims, invariants, behavior specs, edge cases, and evidence.\n  - [`contract-first`](agents/.agents/skills/contract-first/SKILL.md): Approve caller-facing interfaces or shared structure before implementation.\n- Design:\n  - [`specify`](agents/.agents/skills/specify/SKILL.md): Design-partner mode for discovery, tradeoffs, decisions, and design artifacts.\n  - [`domain-modeling`](agents/.agents/skills/domain-modeling/SKILL.md): Data shapes, invariants, state transitions, parsing, and effects.\n  - [`architecture`](agents/.agents/skills/architecture/SKILL.md): Architecture decisions, module boundaries, coupling, layering, and system shape.\n- Correctness and change:\n  - [`code-review`](agents/.agents/skills/code-review/SKILL.md): Review diffs and PRs for bugs, regressions, edge cases, and merge readiness.\n  - [`debugging`](agents/.agents/skills/debugging/SKILL.md): Reproduce symptoms, isolate causes, inspect evidence, and fix bugs.\n  - [`error-handling`](agents/.agents/skills/error-handling/SKILL.md): Error types, propagation, retries, user messages, and recovery.\n  - [`refactoring`](agents/.agents/skills/refactoring/SKILL.md): Behavior-preserving change, tests, and safe rewrites.\n  - [`official-source-check`](agents/.agents/skills/official-source-check/SKILL.md): Check external behavior against official sources.\n- Safety:\n  - [`security`](agents/.agents/skills/security/SKILL.md): Auth, secrets, crypto, input validation, dependency risk, and trust boundaries.\n  - [`database`](agents/.agents/skills/database/SKILL.md): Schemas, migrations, indexes, transactions, query plans, and locking.\n  - [`release`](agents/.agents/skills/release/SKILL.md): Release prep and release-artifact sync, on request or approval.\n- Public surfaces:\n  - [`api`](agents/.agents/skills/api/SKILL.md): REST API contracts: endpoints, fields, evolution, status codes, errors, pagination, idempotency.\n  - [`documentation`](agents/.agents/skills/documentation/SKILL.md): READMEs, ADRs, runbooks, API docs, and comments.\n  - [`ui-design`](agents/.agents/skills/ui-design/SKILL.md): Frontend UI, layouts, components, responsive behavior, and usability.\n  - [`accessibility`](agents/.agents/skills/accessibility/SKILL.md): WCAG, ARIA, keyboard, focus, contrast, and inclusive states.\n- Production quality:\n  - [`async-systems`](agents/.agents/skills/async-systems/SKILL.md): Concurrency, queues, streams, pub/sub, ordering, and backpressure.\n  - [`observability`](agents/.agents/skills/observability/SKILL.md): Logs, metrics, traces, health checks, dashboards, alerts, and SLOs.\n  - [`performance`](agents/.agents/skills/performance/SKILL.md): Profiling, latency, throughput, allocation, caching, and hot paths.\n- Repo workflow:\n  - [`commit`](agents/.agents/skills/commit/SKILL.md): Staging reviewed work, commit splits, and messages.\n  - [`scaffolding`](agents/.agents/skills/scaffolding/SKILL.md): New projects, package setup, quality tooling, CI, and repo structure.\n  - [`git-workflow`](agents/.agents/skills/git-workflow/SKILL.md): Branches, history edits, conflicts, rebases, recovery, and force-push.\n\nGreenfield stack templates live under\n[`scaffolding/references/stacks/`](agents/.agents/skills/scaffolding/references/stacks/).\nShared language defaults are in\n[`language-defaults.md`](agents/.agents/skills/scaffolding/references/language-defaults.md).\n\n## How routing works\n\nConsult routing is **collaboration-aware, quality-driven, and risk-triggered**.\nRisk determines which skills load; working mode determines whether the agent\nshould continue, ask for approval, or stay read-only.\n\nThe working modes are **Direct**, **Guided**, **Design-partner**, and\n**Review-only**. Most implementation work stays in Direct or Guided mode.\n\nConsult is autonomous by default and consultative for significant or hard-to-change\nwork. The agent should get a plan or shape/API sign-off before significant new\ncode (a substantial new module or component, non-trivial logic, or a deliberate\nbehavior change) and before it locks in a caller-facing interface, class or\nlibrary API, project/package/module structure, structural runtime dependency,\ndata model, or boundary that future work will depend on. Local helpers, private\nfile moves, and narrow bug fixes that restore intended behavior should not become\nconsultation gates.\n\nCaller-facing interfaces and shared structure trigger `contract-first`: the\nagent stops at one recommended contract/API/structure and high-level plan, then\nasks for approval before implementation continues.\n\nSee [`workflow`](agents/.agents/skills/workflow/SKILL.md) for the full\nrouting model and [`contract-first`](agents/.agents/skills/contract-first/SKILL.md)\nfor sign-off on interfaces and shared structure.\n\n## Evaluation\n\n[`eval/README.md`](eval/README.md) benchmarks Codex with and without Consult\nagainst shared engineering tasks. It combines deterministic hidden tests with\nLLM-judged engineering maturity, proof quality, simplicity, and risk handling.\n\n## Contributing\n\nSee [`CONTRIBUTING.md`](CONTRIBUTING.md) for project conventions, the skill\nauthoring template, branching rules, checks, and maintenance steps. Skill\nauthoring rules and pack-versioning policy live in\n[`AGENTS.md`](AGENTS.md#skill-anatomy-enforced-by-the-validator).\n\n## License\n\nMIT: see [`LICENSE`](LICENSE). Third-party/adapted extension notices are listed\nin [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md).\n\n## Uninstall\n\nManual install:\n\n```sh\nstow --target=\"$HOME\" -D agents\n```\n\nManual cleanup may still be needed for tool-specific symlinks.\n\nIf you installed the Claude Code plugin, run these from inside Claude Code:\n\n```text\n/plugin uninstall consult@consult\n/plugin marketplace remove consult\n```\n\nFor Codex, remove Consult from the plugin UI or marketplace commands. For Cursor,\ndisable or uninstall **consult** from the marketplace panel (or remove\n`~/.cursor/plugins/local/consult` and any `~/.cursor/plugins/cache/consult` copy). For\nAntigravity, run `agy plugin uninstall consult`.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fkreek%2Fconsult","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fkreek%2Fconsult","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fkreek%2Fconsult/lists"}