Security audit
Security checks for vulnerabilities and agentic risk
This skill has a legitimate audit purpose, but it broadly inspects multi-agent instruction, memory, user, cron, skill, and hook content without clear user scoping or redaction controls.
Install only if you are comfortable with the invoking agent reading multiple agents' instruction files, memories, user preference files, cron prompts, skills, and hooks. Prefer running it on a narrow, user-approved set of agents and avoid copying secrets, personal data, or unrelated prompt content into reports.
references/discovery.md:33Overbroad Collection of Sensitive Multi-Agent State
Skill contains patterns that could indirectly extract system prompts through rephrasing, translation, summarization, or side-channel techniques.
### Step 1 — Discovery
Enumerate every instruction source: all agents, their workspace files (AGENTS.md, SOUL.md, HEARTBEAT.md, MEMORY.md, IDENTITY.md, USER.md), all cron jobs with inline prompts, all installed skills, and any hook scripts.
**→ Detailed procedure: [references/discovery.md](references/discovery.md)**
The text uses a language-specific example, contrasting a SOUL.md specification of Hebrew with AGENTS.md examples in English. Because the policy requires avoiding forced language or locale without user opt-in, this wording can implicitly endorse a locale constraint without stating that user choice or justification is required.