consult — Claude Code skill: what it does, how to install
consult enforces an advisory protocol with strict read-before-cite checks and explicit uncertainty levels.
MIT ★ 539 updated 2026-10-29
npx skills add ZaxbyHub/opencode-swarm --skill consult -a claude-codeWhat consult does
This skill provides runtime execution rules for the architect in consult mode. It starts by looking up prior guidance in .swarm/context.md, identifies one to three focus domains, and contacts the active sme agent serially. Each project phase allows a ceiling of three sme interactions, after which advice gets cached back into context.md.
The protocol mandates reading files before quoting them. Agents cannot mention code lines, file paths, or CLI flags from memory or outdated discussion turns without opening the file in the present revision. When direct verification is impossible during the turn, the agent marks confidence: LOW and lists missing evidence.
Assigning confidence: HIGH to unverified claims violates the protocol directly.
What it helps with
- verify source lines before quoting them in advisory replies
- cap SME agent consultations at three per project phase
- store synthesized advice inside
.swarm/context.md - tag unverified assumptions with
confidence: LOW
Example prompts for the agent
- Run mode consult to inspect current project interfaces with the sme agent
- Review the modified files under consult protocol and verify lines before citing
- Check cached advice in
.swarm/context.mdand consult on new domains
How to install consult
1Skills CLI (Vercel)
npx skills add ZaxbyHub/opencode-swarm --skill consult -a claude-code2Manually: put the files in place
~/.claude/skills/consult/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/consult/
1Skills CLI (Vercel)
npx skills add ZaxbyHub/opencode-swarm --skill consult -a codex2Manually: put the files in place
.agents/skills/consult/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add ZaxbyHub/opencode-swarm --skill consult -a cursor2Manually: put the files in place
.cursor/skills/consult/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install ZaxbyHub/opencode-swarm .claude/skills/consult2Manually: put the files in place
.github/skills/consult/SKILL.md1Skills CLI (Vercel)
npx skills add ZaxbyHub/opencode-swarm --skill consult -a gemini-cli2Platform CLI
gemini skills install https://github.com/ZaxbyHub/opencode-swarm --path .claude/skills/consultManually: put the files in place
.devin/skills/consult/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add ZaxbyHub/opencode-swarm --skill consult -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about consult
How to install consult in Claude Code?
Run npx skills add ZaxbyHub/opencode-swarm --skill consult -a claude-code.
What does the consult skill do?
It guides advisory queries across one to three domains, limits serial SME consultations to three per phase, and prevents stale citations.
How does the protocol handle uncertain claims?
The agent assigns confidence: LOW and specifies which file read, test execution, or documentation check will resolve the question.
Source text
SKILL.md · MIT
--- name: consult description: > Full execution protocol for MODE: CONSULT -- answering advisory questions with bounded evidence and clear uncertainty. metadata: audience: swarm-plugin --- # Consult Protocol This protocol is loaded on demand by the architect runtime. The architect prompt keeps only activation, action, and hard safety constraints; the full execution details live here. ### MODE: CONSULT Check .swarm/context.md for cached guidance first. Identify 1-3 relevant domains from the task requirements. Call the active swarm's sme agent once per domain, serially. Max 3 SME calls per project phase. Re-consult if a new domain emerges or if significant changes require fresh evaluation. Cache guidance in context.md. ### Read-before-cite discipline SME consultation answers are advisory. When the SME (or any agent synthesizing the final answer) cites source code, file paths, line ranges, API behavior, or CLI flags, it MUST read the actual file or tool output in the current revision. Specific anti-patterns to refuse: - Quoting `src/foo.ts:123-145` without reading those lines. - Restating an API signature from earlier conversation history after the source has been edited. - Citing behavior from documentation that pre-dates the current version. For uncertain claims, the SME must mark `confidence: LOW` and identify what additional evidence (a specific file read, a test run, a docs check) would resolve the uncertainty. Do NOT report `confidence: HIGH` for claims that could not be verified against current source this turn.
© ZaxbyHub · MIT · ZaxbyHub/opencode-swarm