agentleFS
tamldtamld/defense-in-depth/.agents/skills/skill-bootstrap-agent/SKILL.md

Bootstrap Agent

Skill4 starsChanged 7 months agoMITView on GitHub

5-phase onboarding wizard for AI agents entering a defense-in-depth project. Self-identify, discover rules, scaffold artifacts, learn to contribute, verify setup.

The file

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 11 are distinctive, and 1 is 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.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Goal: Ensure the project has all required artifacts.

    line 126
  2. Distinctive:

    If your platform has no prebuilt config → create a router file

    line 151
  3. Distinctive:

    Create empty placeholder files to "pass" scaffold check

    line 162
  4. Distinctive:

    Every claim you make must carry an evidence tag:

    line 179
  5. Distinctive:

    When you learn something, record it with this schema:

    line 186
  6. Distinctive:

    Commit without evidence tags on findings

    line 205
  7. Very specific:

    Write lessons without wrongApproach

    line 206
  8. Distinctive:

    Use non-English in code, docs, or commits

    line 207
  9. Distinctive:

    Create hollow artifacts (TODO/TBD = violation)

    line 209
  10. Distinctive:

    If any answer is "No" → return to Phase 2: DISCOVER.

    line 247
  11. Distinctive:

    Start coding before completing all 5 phases

    line 252
  12. Distinctive:

    After completing bootstrap, keep this card handy:

    line 258

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

🚀 Bootstrap Agent — Onboarding Wizard

"A well-onboarded agent is a productive agent. A confused agent is a liability."

This skill guides any AI agent through 5 phases to become a productive contributor in a defense-in-depth governed project.


Overview Flowchart

flowchart TD
    ENTER["Agent enters project"] --> P1["① IDENTIFY\nWho am I?"]
    P1 --> P2["② DISCOVER\nWhat are the rules?"]
    P2 --> P3["③ SCAFFOLD\nWhat's missing?"]
    P3 --> P4["④ CONTRIBUTE\nHow do I add value?"]
    P4 --> P5["⑤ VERIFY\nAm I ready?"]
    P5 -->|"All checks pass"| READY["✅ Ready to work"]
    P5 -->|"Issues found"| P3

Phase 1: IDENTIFY (Who Am I?)

Goal: Establish self-awareness before touching anything.

Steps

  1. Read the project gateway
    Read: AGENTS.md (project root)
    Read: .agents/AGENTS.md (ecosystem map)
  2. Identify your platform
    PlatformConfig FileNative Space
    Gemini CLIGEMINI.md.gemini/
    Claude CodeCLAUDE.md.claude/
    Cursor.cursorrules.cursor/
    OtherCreate a router filePlatform docs
  3. Read your prebuilt config (if exists)
    • These files are cognitive frameworks, not just configs
    • They define how you THINK about this project
  4. Answer these questions internally:
    • What platform am I running on?
    • What are my native capabilities (memory, tools, search)?
    • What can I do that other agents on other platforms can't?

Do Not Do

  • Skip reading AGENTS.md and jump to coding
  • Assume you know the project structure without reading it
  • Modify prebuilt configs without completing all 5 phases

Phase 2: DISCOVER (What Are the Rules?)

Goal: Understand the governance contract before writing any code.

Steps

  1. Read COGNITIVE_TREE.md — the project's philosophical foundation
    Read: .agents/philosophy/COGNITIVE_TREE.md
  2. Read ALL mandatory rules (11 total)
    Read: .agents/rules/rule-consistency.md          (project standards)
    Read: .agents/rules/rule-guard-lifecycle.md       (guard maturity)
    Read: .agents/rules/rule-contribution-workflow.md (how to contribute)
    Read: .agents/rules/rule-evidence-tagging.md      (proof mandate)
    Read: .agents/rules/rule-hitl-enforcement.md      (supreme law)
    Read: .agents/rules/rule-lesson-quality.md        (specificity gate)
    Read: .agents/rules/rule-zero-theater.md          (substance mandate)
    Read: .agents/rules/rule-adaptive-language.md     (language policy)
    Read: .agents/rules/rule-anti-yes-man.md          (brainstorm mandate)
    Read: .agents/rules/rule-agent-workspace.md       (workspace zones)
    Read: .agents/rules/rule-security-continuity.md   (fortress mandate)
  3. Read key contracts
    Read: .agents/contracts/guard-interface.md        (guard API contract)
  4. Read the workflow
    Read: .agents/workflows/procedure-task-execution.md

Key Takeaways Checklist

  • I understand the 3 cognitive branches (Evidence, Mechanism, Growth)
  • I understand HITL is the supreme rule
  • I know what evidence tags are and when to use them
  • I know the lesson schema (wrongApproach is mandatory)
  • I know the consistency rules (TypeScript strict, conventional commits, etc.)
  • I know my private workspace zone vs shared contribution zone

Do Not Do

  • Skim rules without understanding — leads to violations
  • Assume rules from other projects apply here
  • Skip the contracts — they define the API you must respect

Phase 3: SCAFFOLD (What's Missing?)

Goal: Ensure the project has all required artifacts.

Check: defense.config.yml exists

# If missing, create it
npx defense-in-depth init

Check: Required directories

.agents/
  ├── rules/           (11 files minimum)
  ├── workflows/       (≥1 procedure)
  ├── contracts/       (guard interface)
  ├── philosophy/      (COGNITIVE_TREE.md)
  ├── skills/          (this bootstrap skill)
  └── config/          (guards.yml)

Check: Git hooks installed

npx defense-in-depth doctor

Check: Prebuilt config matches your platform

  • If your platform has no prebuilt config → create a router file
  • Router file format:
    You are operating in a defense-in-depth governed project.
    Read AGENTS.md at the project root to understand governance.
    Follow .agents/philosophy/COGNITIVE_TREE.md for mindset.
    Follow .agents/rules/ for mandatory standards.

Do Not Do

  • Skip defense-in-depth doctor — it catches setup issues
  • Create empty placeholder files to "pass" scaffold check
  • Modify existing rules during scaffold phase

Phase 4: CONTRIBUTE (How Do I Add Value?)

Goal: Understand the contribution protocol.

The Contribution Zones

ZoneLocationWhat Goes HereLanguage
Private.gemini/, .claude/, etc.Drafts, reasoning, session stateAny
Sharedsrc/, .agents/, docs/Guards, rules, docs, lessonsEnglish

Evidence Protocol

Every claim you make must carry an evidence tag:

  • [CODE] — verified by reading source code
  • [RUNTIME] — verified by execution
  • [INFER] — inferred from structure
  • [HYPO] — hypothesis, unverified

Lesson Recording

When you learn something, record it with this schema:

{
  scenario: "What happened (specific context)",
  wrongApproach: "What was tried that failed (concrete steps)",
  correctApproach: "What actually fixed it (concrete steps)",
  insight: "The generalizable takeaway",
  searchTerms: ["term1", "term2"],  // ≥2 terms
  evidenceLevel: "RUNTIME"          // How you verified
}

Commit Format

<type>(<scope>): <description>

Types: feat, fix, chore, docs, refactor, test

Do Not Do

  • Commit without evidence tags on findings
  • Write lessons without wrongApproach
  • Use non-English in code, docs, or commits
  • Auto-merge without meeting ALL criteria
  • Create hollow artifacts (TODO/TBD = violation)

Phase 5: VERIFY (Am I Ready?)

Goal: Confirm setup and readiness.

Readiness Checklist

# 1. Project health
npx defense-in-depth doctor

# 2. All guards pass on current code
npx defense-in-depth verify

# 3. TypeScript compiles
npx tsc --noEmit

# 4. Tests pass (if test suite exists)
npm test

Self-Assessment

Answer these honestly:

QuestionExpected Answer
Can I explain the 3 cognitive branches?Yes — Evidence, Mechanism, Growth
Do I know the supreme rule?Yes — HITL
Can I tag evidence correctly?Yes — [CODE], [RUNTIME], [INFER], [HYPO]
Do I know where my private workspace is?Yes — platform-specific directory
Do I know the lesson schema?Yes — including wrongApproach
Can I write a conventional commit?Yes — type(scope): description
Do I understand what theater is?Yes — hollow artifacts

If any answer is "No" → return to Phase 2: DISCOVER.

Do Not Do

  • Claim readiness without running the verification commands
  • Skip self-assessment — overconfidence leads to violations
  • Start coding before completing all 5 phases

Quick Reference Card

After completing bootstrap, keep this card handy:

┌────────────────────────────────────────┐
│  🛡️ defense-in-depth Quick Reference   │
├────────────────────────────────────────┤
│  Supreme Rule: HITL                   │
│  Evidence: [CODE] [RUNTIME] [INFER]   │
│  Language: Chat=Any, Docs=English     │
│  Commits: type(scope): description    │
│  Theater: TODO/TBD = VIOLATION        │
│  Lessons: wrongApproach = MANDATORY   │
│  Guards: Observe + Report, NO mutate  │
│  Verify: npx defense-in-depth verify   │
│  Health: npx defense-in-depth doctor   │
└────────────────────────────────────────┘

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.