agentsclimarketplace

Agent redline

Skill rore/agent-redline/dist/agent-redline

Agent governance as a skill. Stops AI agents from quietly turning local fixes into architecture debt.

Install
npx -y skills add rore/agent-redline --skill agent-redline

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when setting up agent governance for a repo, or when working in a repo that contains agent-policy.yaml. Classifies structural risk and routes human attention to red-zone changes.

SKILL.md

3.0 KB, 681 tokens by cl100k_base, as published. Nobody here has run it

agent-redline

You classify changes before editing. Deterministic CI checks (a boundary-rule backend, the reporter) catch what you miss.

Vocabulary

TermMeaning
Red zoneChanges are dangerous: contracts, modeling, architecture, security, persistence, shared behavior.
Blue zoneAutonomous work fine: isolated, replaceable, strongly testable, low blast-radius.
Gray zoneUnclassified. Cautious by default.
WatchAdditive tag (not a zone): paths surfaced in the PR comment regardless of red/blue/gray classification. No checkpoint, no gate — visibility only.
Boundary ruleDeterministic dependency rule (X must not import Y). Enforced by a backend (e.g., ArchUnit).
CheckpointRequired human attention; satisfied by a label or CODEOWNER approval. Types: architecture-review, api-review, persistence-review, security-review, ops-review.

Pick a mode

  1. agent-policy.yaml exists in the repo root → read operating-mode.md. Everyday work.
  2. The user asked you to set up agent-redline → read bootstrap-mode.md. One-time setup.
  3. Neither → this skill is not relevant.

Read only the file for the mode that applies.

Principles (non-negotiable)

  1. Classify before editing. Decide blue / red / gray before you touch a file; note any watch paths to surface in the PR.
  2. Refuse boundary shortcuts. If a change would create a forbidden dependency, fix the structure or escalate. Do not suppress, do not modify the architecture-test files, do not launder the import.
  3. Architecture-test files are red. Any change to the boundary-rule backend's definition files (e.g., ArchUnit test classes) requires architecture-review, regardless of what the policy says.
  4. Never auto-commit CI changes. Workflows, branch protection, CODEOWNERS — propose, don't commit.
  5. Default conservative on uncertainty. Gray > blue, red > gray, boundary risk > everything.
  6. No slop. Tight PR descriptions, tight code comments. What's not obvious from the code, and what it does.

Decision priority

When ambiguous:

  1. The repo's agent-policy.yaml
  2. The repo's AGENTS.md / CLAUDE.md / GEMINI.md
  3. The active language extension's operating.md (if present)
  4. These principles
  5. Ask the developer

Do not invent rules the policy doesn't state.

Resources (load on demand)

  • operating-mode.md — read when entering operating mode
  • bootstrap-mode.md — read when entering bootstrap mode
  • Active extension's files in extensions/<name>/ — profile.md and scaffold.md during bootstrap; operating.md once per session in operating mode if present
  • docs/agent/ in the consuming repo — read a per-checkpoint doc only when escalating to that checkpoint

What ships with it: 36 files

294.2 KB alongside SKILL.md, 5 of them executable

assets/

extensions/

references/

scripts/

Gives 0 of the 12 instructions most context ai engineering skills give in 681 tokens

Counted across 1,193 of the 1,976 authors here whose files we hold, read 2026-08-07

  • Dispatch a fresh implementer subagent per taskin 48 of 1193, across 19 files
  • Dispatch a final code reviewer after all tasksin 33 of 1193, across 8 files
  • Provide full task text to the subagentin 30 of 1193, across 9 files
  • Review spec compliance before code qualityin 27 of 1193, across 10 files
  • Make the hook script executablein 26 of 1193, across 8 files
  • Re-snapshot after navigation or DOM changesin 25 of 1193, across 19 files
  • Read files before editing themin 22 of 1193, across 11 files
  • Answer subagent questions before proceedingin 22 of 1193, across 7 files
  • Mark task complete in TodoWrite after approvalin 22 of 1193, across 6 files
  • Merge hook into existing settingsin 21 of 1193, across 3 files
  • Ask if installation is global or projectin 20 of 1193, across 2 files
  • Copy the hook script to target locationin 20 of 1193, across 2 files

Said here and by no other author read

  • refuse boundary shortcuts
  • treat architecture-test files as red zone
  • escalate or fix when facing boundary violations
  • default to conservative classification on uncertainty
  • never auto-commit CI configuration changes
  • write tight PR descriptions and code comments

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,984. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.