hashgraph-online/awesome-codex-plugins / plugins/boshu2/agentops/skills/security
security
FlaggedReview code for security problems; scan for vulnerabilities, secrets, dependency and prompt risks. Use when: asked whether code is safe to ship, even one small handler.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 10.1 KB |
| references/agentops-redteam-pack.json | config | 6.4 KB |
| references/attacker-path-evidence.md | markdown | 4.7 KB |
| references/owasp-checklist.md | markdown | 6.7 KB |
| references/policy-example.json | config | 525 B |
| references/security-suite-runbook.md | markdown | 8.4 KB |
| scripts/prompt_redteam.pyexecutable | script | 11.3 KB |
| scripts/security_suite.pyexecutable | script | 32.7 KB |
| scripts/validate.shexecutable | script | 5.4 KB |
scripts/validate.sh
View on GitHub#!/usr/bin/env bash
set -euo pipefail
# pwd -P: this skill is invoked through a symlink (~/.claude/skills/security ->
# the checkout); a logical pwd would resolve ../.. against the symlink's parent
# (.claude), so the repo-surface probe below would silently miss AGENTS.md and
# skip the behavioral redteam instead of running it.
SKILL_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd -P)"
SKILL="$SKILL_DIR/SKILL.md"
REPO_ROOT="$(cd "$SKILL_DIR/../.." && pwd -P)"
[[ -s "$SKILL" ]]
grep -q '^name: security$' "$SKILL"
# effects must be DECLARED, not the copied-default empty array. This gate used
# to pin `effects: []`; that pin mechanically enforced a false contract (the
# skill writes scan artifacts). The check is now inverted: an empty effects
# array fails.
if grep -q '^ effects: \[\]$' "$SKILL"; then
echo 'security effects must be declared (the skill writes scan artifacts); effects: [] is a false contract' >&2
exit 1
fi
[[ "$(awk '/^---$/{n++;next} n==2 && /^## /{print;exit}' "$SKILL")" == "## Critical Constraints" ]]
grep -Fq '**Artifact directory:**' "$SKILL"
grep -Fq '**Validator command:**' "$SKILL"
grep -Fq 'report stops after evidence' "$SKILL"
if grep -Eiq 'AUTO-REDO|ONE-HELPER|HELPER-ESCALATE|ao (pawl|land)|next_action' "$SKILL"; then
echo 'security contract contains retired lifecycle vocabulary' >&2
exit 1
fi
# The skill-local security-gate.sh duplicate was unrunnable: its REPO_ROOT
# resolved to skills/security, so it looked for a nonexistent skill-local
# toolchain-validate.sh and exited 1. The canonical gate is the repo-root
# scripts/security-gate.sh (documented in SKILL.md). Guard against the
# duplicate coming back.
if [[ -e "$SKILL_DIR/scripts/security-gate.sh" ]]; then
echo 'skill-local scripts/security-gate.sh duplicate is back — it is unrunnable; use the repo-root gate' >&2
exit 1
fi
[[ -s "$SKILL_DIR/references/policy-example.json" ]]
[[ -s "$SKILL_DIR/references/agentops-redteam-pack.json" ]]
[[ -s "$SKILL_DIR/references/security-suite-runbook.md" ]]
# Syntax-check the shipped Python without writing __pycache__/*.pyc into the
# package (py_compile writes the default cfile even with cfile=None); ast.parse
# validates syntax and writes nothing.
python3 - "$SKILL_DIR/scripts/security_suite.py" "$SKILL_DIR/scripts/prompt_redteam.py" <<'PY'
import ast
import sys
for path in sys.argv[1:]:
with open(path, encoding="utf-8") as fh:
ast.parse(fh.read(), filename=path)
PY
python3 -c 'import json, pathlib, sys; root=pathlib.Path(sys.argv[1]); [json.loads((root/name).read_text()) for name in ("policy-example.json", "agentops-redteam-pack.json")]' "$SKILL_DIR/references"
# Seeded suite cases: Cobra grouped help must yield commands (it yielded none,
# so the documented ao policy example failed), the starter deny list must match
# the remote side of an lsof row, and lsof must AND its selectors (-a) so other
# processes' sockets are not recorded.
python3 -I - "$SKILL_DIR/scripts/security_suite.py" "$SKILL_DIR/references/policy-example.json" <<'PY'
import importlib.util, json, re, sys
sys.dont_write_bytecode = True
spec = importlib.util.spec_from_file_location("security_suite", sys.argv[1])
suite = importlib.util.module_from_spec(spec)
sys.modules["security_suite"] = suite
spec.loader.exec_module(suite)
grouped = ("Usage:\n ao [command]\n\nGetting Started:\n init Create\n\n"
"Core Commands:\n status Show\n\nFlags:\n -h, --help help\n\n"
'Use "ao [command] --help" for more information about a command.\n')
assert suite._parse_subcommands(grouped) == ["init", "status"], "grouped help parse"
assert suite._parse_subcommands("Usage: t FILE\n\nNotes:\n Read once.\n") == [], "prose parsed as commands"
deny = json.load(open(sys.argv[2]))["deny_network_endpoint_patterns"]
assert any(re.search(p, "TCP 203.0.113.5:5000->10.0.0.1:443") for p in deny), "remote private missed"
assert not any(re.search(p, "TCP 192.168.1.5:5000->203.0.113.9:443") for p in deny), "local LAN flagged"
assert '"lsof", "-nP", "-a", "-i", "-p"' in open(sys.argv[1]).read(), "lsof selectors not ANDed"
PY
# Behavioral redteam: run the attack pack against the live repo surfaces and
# require verdict PASS. This is what keeps the pack honest — a stale target glob
# or pattern (the "fails against its own tree" defect) turns this red. It is
# only meaningful when the governance surfaces the pack targets are present, so
# it is gated on the real repo; an isolated skill copy or standalone install
# (which lacks AGENTS.md and docs/) skips it with a disclosed note rather than
# failing on absent, unscannable surfaces.
if [[ -f "$REPO_ROOT/AGENTS.md" && -f "$REPO_ROOT/docs/CI-CD.md" ]]; then
redteam_out="$(mktemp -d "${TMPDIR:-/tmp}/security-redteam.XXXXXX")"
trap 'rm -rf "$redteam_out"' EXIT
if python3 "$SKILL_DIR/scripts/prompt_redteam.py" scan \
--repo-root "$REPO_ROOT" \
--pack-file "$SKILL_DIR/references/agentops-redteam-pack.json" \
--out-dir "$redteam_out" >/dev/null 2>&1; then
echo "security redteam: PASS (attack pack holds against the live tree)"
else
echo 'security redteam FAILED against the live tree — a target glob or pattern is stale, or a control regressed' >&2
jq -r '.failed_cases[]?' "$redteam_out/redteam/redteam-results.json" 2>/dev/null \
| sed 's/^/ failed case: /' >&2 || true
exit 1
fi
else
echo "security redteam: SKIP (repo governance surfaces absent; not the AgentOps repo)"
fi
echo "security contract: PASS"
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 465k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 465k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 465k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 192k | Updated | Claude CodeCodex | Verified | Install |