Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsCommunityBlog
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Gibson Setup

ASecurity

Wire a repository for autonomous fleet development: install the Gibson guardrail bundle (AGENTS section, CI gates, labels, branch protection, kill switch) and connect the Devin cloud supervisor as merge captain. Consumes gibson-audit's gap list; idempotent. Third stage of the /gibson pipeline; also standalone ('set up this repo for the fleet', 'gibson-ify this repo', 'install the guardrails').

26 stars
0 votes
0 copies
0 views
Added 9/28/2026
ai-agentsrustgonoderailsapisecurity

Works with

api

Security Analysis

A100/100

Scanned 9/28/2026

Install to Claude Code

$npx -y skills add The-AIE/the-gibson --skill gibson-setup --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Gibson Setup?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Gibson Setup
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/the-aie-gibson-setup/badge)](https://www.skillsdirectory.com/skills/the-aie-gibson-setup)

More formats (shields.io, HTML) on the badges page.

Files
SKILL.md
---
name: gibson-setup
description: "Wire a repository for autonomous fleet development: install the Gibson guardrail bundle (AGENTS section, CI gates, labels, branch protection, kill switch) and connect the Devin cloud supervisor as merge captain. Consumes gibson-audit's gap list; idempotent. Third stage of the /gibson pipeline; also standalone ('set up this repo for the fleet', 'gibson-ify this repo', 'install the guardrails')."
---

# gibson-setup — teeth, not just prose

An `AGENTS.md` alone is policy in prose; agents drift from it and nothing
notices. This skill installs the ENFORCEMENT bundle so the rules are checked by
machines. Source of truth: `templates/target-repo/` in the Gibson clone (currently thin —
one file — plus the `ci/` workflows; be honest about that). When a target repo
needs something the templates lack, write it for the target, then contribute
the generalized version back to the Gibson **via a normal worktree branch + PR
(Law 3: never edit the canonical clone directly)** so the next repo gets it
for free.

## Install checklist (drive from the audit's gap list; skip what exists)

1. **AGENTS section** — append `templates/target-repo/AGENTS-section.md` to the
   target's `AGENTS.md` (create if absent). Never overwrite existing content.
2. **CI gates** — test/lint/typecheck/build on every PR, as required checks.
   Reuse the repo's existing CI if present; add the missing jobs only.
   **Verify every referenced script actually exists in the TARGET repo before
   installing any workflow from `ci/`.** `ci/gibson-gate.yml` runs
   `node scripts/check-active-work.mjs` (claim isolation, issue #55) and the
   protected **test-integrity** four-job path (issue #70) that grades with the
   merge-base copy of `scripts/test-integrity.mjs` only. Vendor **both** scripts
   into the target alongside the workflow. Calibrate `<generate>`/`<typecheck>`/
   `<lint>`/`<test>`/`<build>` **and** set `TEST_COMMAND` on both
   `test-integrity-base` and `test-integrity-head` to the same literal as the
   gate test step — never read PR-head `.agents/gate.json` for capture.
   Claim isolation needs `fetch-depth: 0` on the green-gate job.
   Anything the workflow references that you have not vendored must be vendored,
   stubbed, or trimmed — a missing script fails every PR with module-not-found.
   **Install ≠ protect:** copying `gibson-gate.yml` is inert until the owner
   makes `test-integrity` a required check and runs the #68 canaries (no-change
   pass, deletion/skip/hostile-helper/failing-base/missing-artifact fails, exact
   waiver pass, workflow-file modification protection). Branch protection,
   rulesets, fork policy, merge-queue policy, and credentials are owner Ask
   Contract items under #68 — never agent-applied. Merge queues stay off until
   equivalent `merge_group` support is implemented and canaried.
3. **Risk classifier** — Tier-C auto-labeling for ALL Law 7 categories: money,
   auth, consent/PII, security boundaries, production data (schema/migrations
   included). Start from the Gibson's own `ci/` workflows; where a needed gate
   has no template yet, port the concept from a proven external implementation
   (e.g. conference-os's risk classifier — lives in that repo, not this one)
   and contribute the generalized version back via PR (see rule below).
4. **Review-evidence gate** — PRs need independent review evidence before
   merge; owner attestation path for humans; trusted-provider list for bots
   (Devin's App belongs here IF the owner approves — that file is a security
   boundary, adding to it is always an explicit owner decision, never implied).
   **No local template for this exists yet** — port it from a proven external
   implementation and contribute the generalized version back (rule below).
   Until installed, say plainly the repo has no machine-checked review gate.
5. **Labels + kill switch** — `agent-claimed`, tier labels, and `gibson-halt`.
   The permanent stop is the `gibson/HALT` file; `GIBSON_HALT=1` is also checked
   unconditionally every iteration. The `gibson-halt` label is only a SOFT cue —
   `loop.sh` honors it when `gh` is installed and authenticated on the machine
   running the loop, and silently ignores it otherwise. Either also install the
   small translation workflow (label applied → touch `gibson/HALT`) or state
   plainly that on a box without `gh` the label stops nothing and humans must
   create the HALT file to actually stop the fleet.
6. **Branch protection** — required checks + no force-push on the default
   branch. Needs owner-level auth; if unavailable, emit the exact settings as
   an Ask Contract item instead of failing silently.
7. **DCO/sign-off** convention if the repo wants it.
8. **Devin supervisor wiring** — `scripts/devin-supervisor.sh ensure --repo <path>`
   **creates a billed cloud session if none exists (ACUs = real money): this is
   an owner-gated step, always. Ask Contract first, run only on an explicit
   yes** (needs `DEVIN_API_KEY`; set `DEVIN_MAX_ACU` as the cap). Optional
   webhook wake per `adapters/devin/README.md`. Devin = merge captain: reviews
   finished branches and owns PRs; it merges only in the explicit `--merge`
   handoff mode (docs/22) — the default leaves every merge to a human. Tier C
   ends at the human gate in every mode (Law 7).

## Rules

- **Idempotent**: re-running against an already-wired repo changes nothing and
  says so. Diff-then-write, never blind-write.
- **Right-sized, never under-gated**: install the lean baseline everywhere; add
  the heavy machinery where the audit found risk surfaces. But Law 7 is not
  profile-dependent — if ANY money/auth/consent/PII/security/prod-data surface
  exists, its human merge gate gets wired regardless of chosen profile. A
  static brochure site with a contact form still has a PII surface.
- Every owner-credential step (branch protection, App installs, API keys) is an
  Ask Contract item — batched, never a drip.
- Commit the wiring via a normal PR to the target repo, reviewed cross-vendor
  like any other change. Setup is not exempt from the rules it installs.

Attribution

The-AIEThe-AIE
View sourceMore from The-AIE →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Ultra-compressed communication mode that cuts output tokens while keeping technical accuracy. Levels: lite, full, ultra and the wenyan variants. Use for /caveman, "caveman mode", "talk like caveman", "be brief" or "less tokens".

1074701 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

695601 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3351 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

691 votes

math-skill

A comprehensive mathematical reasoning skill for AI assistants — handles arithmetic to research-level problems with rigorous step-by-step reasoning, systematic verification, and transparent uncertainty handling

381 votes
View all in ai-agents →