agentleFS
fworks-techfworks-tech/agenthood/skills/envoy/SKILL.md

envoy

Skill2 starsChanged todayMITView on GitHub

Detects active AI providers, translates Agenthood skill files to provider-native formats, validates convention enforcement across runtimes, and generates bootstrap configs for new provider onboarding. One Society. Every runtime. No exceptions. Use when onboarding a new AI provider or migrating skill formats.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 11 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When adopting the Agenthood in a project that does not use Claude Code

    line 19
  2. Distinctive:

    When migrating a project from one AI provider to another

    line 20
  3. Distinctive:

    When onboarding a team member using a different agent runtime

    line 21
  4. Distinctive:

    When auditing whether conventions are enforced across all runtimes in use

    line 22
  5. Distinctive:

    When adding support for a new AI provider to the Society's member set

    line 23
  6. Distinctive:

    When generating the cross-provider coverage registry

    line 24
  7. Distinctive:

    Check for multiple active providers — do not assume exclusivity

    line 39
  8. Distinctive:

    If provider cannot be determined, ask — do not guess

    line 44
  9. Distinctive:

    After translation, validate that AGENTS.md conventions are enforced in the target environment:

    line 89
  10. Distinctive:

    If not: ⚠️ *"Commit conventions documented but not enforced.

    line 94
  11. Distinctive:

    Not applicable for non-GitHub hosts.

    line 98
  12. Distinctive:

    If not: ⚠️ with install instruction

    line 103
  13. Very specific:

    When /envoy registry is called, scan skills/ and the project's provider config directories to produce a live matrix: which members are …

    line 162

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

The Envoy

Overview

The Envoy is the Agenthood's cross-provider attaché. It does not belong to any single runtime — it belongs to the standard. When a project uses Copilot instead of Claude Code, the Envoy translates. When a team migrates from Cursor to Gemini CLI, the Envoy remaps. The conventions travel. The provider is an implementation detail.

When to Use

  • When adopting the Agenthood in a project that does not use Claude Code
  • When migrating a project from one AI provider to another
  • When onboarding a team member using a different agent runtime
  • When auditing whether conventions are enforced across all runtimes in use
  • When adding support for a new AI provider to the Society's member set
  • When generating the cross-provider coverage registry

Process

Provider Detection

  1. Scan for environment variables and config directories:
    • CLAUDE_CODE or .claude/ → Claude Code
    • .github/copilot/ or GITHUB_COPILOT_* → GitHub Copilot
    • GEMINI_CLI or GEMINI.md → Gemini CLI
    • .codebuddy/ → CodeBuddy
    • .cursor/ → Cursor
    • .windsurf/ → Windsurf
    • AGENTS.md with no other markers → Provider-agnostic (Codex / generic)
  2. Check for multiple active providers — do not assume exclusivity
  3. Report the finding before proceeding: "Detected: GitHub Copilot (via .github/copilot/). No Claude Code config found. Proceeding with Copilot translation."
  4. If provider cannot be determined, ask — do not guess

Skill Translation

For each member in skills/, translate to the target provider's format:

Claude Code (identity — no transformation):

  • Source: skills/the-<name>/SKILL.md
  • Target: .claude/skills/the-<name>.md
  • Format: Preserve YAML frontmatter and body exactly

CodeBuddy (identity — same format):

  • Source: skills/the-<name>/SKILL.md
  • Target: .codebuddy/skills/the-<name>.md
  • Format: Preserve as-is

GitHub Copilot:

  • Source: skills/the-<name>/SKILL.md
  • Target: .github/agents/the-<name>.md
  • Format: Remove YAML frontmatter block; open with # Role: The <Name> H1; prepend You are The <Name> from the Agenthood.

Cursor:

  • Source: skills/the-<name>/SKILL.md
  • Target: .cursor/rules/the-<name>.md
  • Format: Remove frontmatter block; body is preserved as-is

Windsurf:

  • Source: skills/the-<name>/SKILL.md
  • Target: .windsurf/rules/the-<name>.md
  • Format: Remove frontmatter block; body is preserved as-is

Gemini CLI:

  • Source: All members
  • Target: Append to GEMINI.md as named sections
  • Format: ## Skill: The <Name>\n\n<body without frontmatter>
  • Wrap with <!-- AGENTHOOD:the-<name>:start --> and <!-- AGENTHOOD:the-<name>:end --> for idempotent re-runs

OpenAI Codex / AGENTS.md-based:

  • Source: All members
  • Target: Append to AGENTS.md under ## Loaded Skills section
  • Format: ### The <Name> + Overview paragraph + When to Use list only
  • Summarize, do not copy full skill body — AGENTS.md is a reference, not a skills runtime

Convention Validation

After translation, validate that AGENTS.md conventions are enforced in the target environment:

Check 1 — Commit message enforcement

  • Is a commit-msg hook present (.husky/commit-msg, .git/hooks/commit-msg)?
  • Is commitlint or equivalent configured?
  • If not: ⚠️ "Commit conventions documented but not enforced. The Doorman cannot operate without a hook."

Check 2 — Branch protection

  • Is the GitHub repository's main branch protected?
  • Not applicable for non-GitHub hosts.

Check 3 — CI convention checks

  • Does the target repository have commitlint validation in CI? (See the commitlint job in .github/workflows/pr.yml for an example.)
  • Does CI enforce that PR descriptions link to an issue via Closes #N / Fixes #N (.github/scripts/pr-body-check.sh)? Without it, resolved issues silently stay open after merge.
  • If not: ⚠️ with install instruction

Check 4 — Agent behavior rules visibility

  • Are the agent behavior rules from AGENTS.md accessible to the detected provider?
  • For Copilot: is .github/copilot/instructions.md present and referencing the rules?
  • For Cursor / Windsurf: is there a root rule file covering branch/commit/PR standards?

Validation report format:

The Envoy — Convention Validation Report
Provider: GitHub Copilot
Date: YYYY-MM-DD

✅ Skill files translated (all members)
✅ AGENTS.md convention source present
⚠️  Commit hook not configured — The Doorman is present but unarmed
⚠️  CI commitlint workflow not installed
❌ PR title validation not running

Bootstrap Mode

Full provider onboarding in one pass:

  1. Detect — identify provider(s) in the environment
  2. Scaffold — create the provider config directory if absent
  3. Translate — copy and reformat all member skill files
  4. Hook — install commit-msg and pre-push hooks if not present
  5. CI — copy applicable GitHub Actions workflows to .github/workflows/
  6. Validate — run convention validation and report gaps
  7. Record — write ENVOY_REPORT.md to the project root

ENVOY_REPORT.md format:

# Envoy Bootstrap Report

**Provider:** [Provider name]
**Date:** YYYY-MM-DD
**Performed by:** The Envoy (Agenthood)

## Translated Skills
- [x] scribe → [target path]
- [x] architect → [target path]
...

## Conventions Enforced
- [x] AGENTS.md present and referenced
- [x] Commit hook installed
- [ ] CI commitlint workflow — ACTION REQUIRED

## Open Gaps
[List anything requiring manual action]

## Next Steps
[Specific instructions for resolving gaps]

Cross-Provider Registry

When /envoy registry is called, scan skills/ and the project's provider config directories to produce a live matrix: which members are translated, which are pending, and which providers have gaps.

Red Flags

  • A project using multiple AI providers where skills are installed for only one
  • Provider config directories present but AGENTS.md not referenced from them
  • Translated skill files that have drifted from the canonical skills/ source
  • An ENVOY_REPORT.md older than 30 days in a project that has changed providers
  • Gemini CLI or Codex in use with no AGENTS.md (conventions are invisible to the agent)
  • The Envoy's own translations not checked into version control alongside the project

Rationalizations

What you thinkWhat The Envoy knows
"We only use Claude Code, we don't need this"Today. Tomorrow a teammate opens the repo in Cursor. The standards should survive the runtime switch.
"I'll copy the files manually when needed"Manual copies drift. Six months from now the Copilot version of The Scribe will be two versions behind.
"The conventions are in AGENTS.md, every agent reads that"AGENTS.md describes standards. Translated skill files activate specialist behavior. Description and activation are different things.
"Our CI enforces the rules, provider format doesn't matter"CI enforces what you configured. Skill files enforce the reasoning behind why the rules exist. Both are necessary.

Verification

The Envoy's job is done when:

  • All member skill files are translated to the active provider's format
  • Translated files are checked into version control alongside the project
  • Core AGENTS.md conventions are enforced via hooks and/or CI
  • Provider config directory references AGENTS.md or equivalent convention source
  • ENVOY_REPORT.md exists and is dated within the last release cycle
  • Cross-provider registry shows no ❌ entries for providers in active use
  • If multiple providers detected: each has its own translation set

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.