review-like-a-red-team

Warns only (observe): flags files that make a dev tool run code on open, clone or shell entry -- agent hooks, helpers, env/BASE_URL overrides, auto-approve; VS Code folderOpen tasks, trust off, repo executables; devcontainer initializeCommand; .envrc, mise, husky, lefthook, pre-commit; gitconfig exec keys, gitattributes drivers, unsafe .gitmodules. Commit, agent write, turn's end; additions only; unreadable configs refused. Friction: misses interpreted code and unlisted files. [Session-enforced via PreToolUse and Stop hooks; needs git and a Python 3.11+. PreToolUse judges the file a tool call would write; Stop re-reads what the turn actually left on disk, so a file written through a shell heredoc is judged too. With no working Python the hook exits 2; without git, fail-open clients allow silently. A gate that cannot reach a decision refuses rather than allowing one it never judged. Enforcement at every commit and in CI still needs chock installed in the repo.]

1 skill · hooks

Blocks: parses agent permission configs (.claude/settings*, .codex, .gemini, .vscode, .cursor/cli.json, opencode, .aider, .continue) and flags added bare or wildcard allows (Bash, curl/rm/sudo/git push, WebFetch, Write/Edit globs, mcp__*), removed deny entries, bypass/auto modes, codex never+danger, yolo, autoAccept, yes-always, regex-all VS Code approve, Claude project env API-endpoint overrides and enableAllProjectMcpServers. Misses: other MCP configs, user-level settings, scripts, Read. [Session-enforced via PreToolUse and Stop hooks; needs git and a Python 3.11+. PreToolUse judges the file a tool call would write; Stop re-reads what the turn actually left on disk, so a file written through a shell heredoc is judged too. With no working Python the hook exits 2; without git, fail-open clients allow silently. A gate that cannot reach a decision refuses rather than allowing one it never judged. Enforcement at every commit and in CI still needs chock installed in the repo.]

1 skill · hooks

Best effort, ask only. Asks a person before a shell command switches the identity later commands act as: aws --profile, AWS_PROFILE= or AWS keys (also $env:), sts assume-role, configure set of a key; gcloud account, configuration, login or impersonation; az account set, az login as a principal or identity; kubectl use-context or set-credentials; writes to ~/.aws, ~/.config/gcloud, ~/.azure. Names in .chock/credential-allowlist.json pass. Misses: scripts, SDKs, env files, credential_process. [Session-enforced via a PreToolUse hook; needs git, a usable bash and a Python 3.11+ (python3, python or py, whichever actually runs; the Windows Store stub is skipped). With no working Python the hook refuses (exit 2); without git or bash, fail-open clients allow silently and fail-closed clients refuse matched commands. If the guard itself crashes or times out, the hook asks for confirmation rather than allowing silently.]

1 skill · hooks

Best-effort guard: refuses a download (curl, wget, iwr; ssh host cmd, scp/rsync/aws s3 cp to stdout) wired into a shell or interpreter (sh..fish, python, perl, ruby, node, php, lua, pwsh, deno, bun, su, source, eval, iex) by pipe, substitution or here-string, also inside bash -c/ssh/docker exec bodies, or downloaded then run with no checksum or signature step. nc/socat/one-liner reads ask. Misses: aliases, functions, outside variables, encoded text, renamed downloads. Friction only. [Session-enforced via a PreToolUse hook; needs git, a usable bash and a Python 3.11+ (python3, python or py, whichever actually runs; the Windows Store stub is skipped). With no working Python the hook refuses (exit 2); without git or bash, fail-open clients allow silently and fail-closed clients refuse matched commands. If the guard itself crashes or times out, the hook asks for confirmation rather than allowing silently.]

1 skill · hooks

Best-effort, on parsed commands (echo ignored); one verdict table shared with rtk-dangerous-actions-blocker. Blocks recursive deletes of root, home, abs paths or .git; mv/chmod/chown -R of system dirs; dd to devices, mkfs; lockouts; force, delete, mirror pushes; history rewrites; switch --discard; IaC destroy; cloud deletes; DB drops. Asks: rm -rf off a safe list, one named kubectl delete, bare lease, stash drop, -auto-approve, drain, prunes, API deletes. Dry runs pass. Pre-push: no non-ff. [Session-enforced via a PreToolUse hook; needs git, a usable bash and a Python 3.11+ (python3, python or py, whichever actually runs; the Windows Store stub is skipped). With no working Python the hook refuses (exit 2); without git or bash, fail-open clients allow silently and fail-closed clients refuse matched commands. If the guard itself crashes or times out, the hook asks for confirmation rather than allowing silently.]

1 skill · hooks

Warns (observe rollout) when a line added to a script, Dockerfile, Makefile, CI workflow or package.json wires a download into a code runner: curl/wget/aria2c/lynx/iwr/irm piped into a shell, python/perl/ruby/node/php/lua on stdin, pwsh or iex; bash -c, eval, source or a here-string of a $(...) or <(...) download; Dockerfile ADD of a URL without --checksum. One line at a time: continuation lines, variables, unicode-escaped JSON, split or quoted command names, a download saved to a file and run on a later step, and interpreters or fetchers outside the lists are missed. Never refuses. Friction only. [Session-enforced via PreToolUse and Stop hooks; needs git and a Python 3.11+. PreToolUse judges the file a tool call would write; Stop re-reads what the turn actually left on disk, so a file written through a shell heredoc is judged too. With no working Python the hook exits 2; without git, fail-open clients allow silently. A gate that cannot reach a decision refuses rather than allowing one it never judged. Enforcement at every commit and in CI still needs chock installed in the repo.]

1 skill · hooks

Harness-native ECC operator layer - 71 agents, 314 skills, 98 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses

200 skills · 98 commands · 71 agents · 1 MCP server · hooks

Easily create custom hooks to prevent unwanted behaviors by analyzing conversation patterns or from explicit instructions. Define rules via simple markdown files.

1 skill · 4 commands · 1 agent · hooks

Project Anatole: an optional, learning watchdog for unattended agents (OWASP-mapped rules, a per-project baseline, learn / notify / enforce modes). Default off.

Guided modernization for any legacy codebase: start with /modernize, get an assessment, an interactive map, the business rules mined from the code, and a plan you approve, then an upgrade, a rewrite or a rebuild with independent proof that the new code behaves like the old. Includes security hardening, a one-page report, specialist agents, and an optional live progress pane (early access).

13 commands · 8 agents · hooks

Independent auditing for AI agents, open source. Claude discovers your setup, confirms the cost, and audits your agent against the one question other tools skip: is it doing the job it's supposed to do, given your business rules and org structure? Test any model (Anthropic, OpenAI, Gemini, Azure, Bedrock, …) graded by self, one independent judge, or a cross-vendor panel; each provider is billed to its own account via a key in your settings, never the command line. Results render as an interactive artifact (with a static-report fallback).

1 skill · hooks

Parses onboarding docs, runs the rules engine, flags gaps

3 skills · 1 agent