update-agent-config
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a diagnostic shell script (
scripts/audit-agents-md.sh) to inspect the repository's filesystem. This script is read-only and follows best practices by pruning build and dependency directories (e.g., node_modules, .git, venv) to avoid unnecessary noise. - [COMMAND_EXECUTION]: The skill instructs the agent to perform standard development tasks using Git, including
git logfor churn analysis andgit commit/pushfor persisting documentation updates. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted documentation files (AGENTS.md) which presents a potential attack surface. However, the skill's core 'falsifiable-claim' discipline serves as a defense-in-depth measure. 1. Ingestion points: Reads AGENTS.md and source code files. 2. Boundary markers: The skill enforces an 'audit-first' workflow where claims must be verified against the filesystem. 3. Capability inventory: Uses bash, git, grep, sed, and wc. 4. Sanitization: All claims are cross-referenced with the ground truth of the source code before the agent acts on them.
- [SAFE]: The skill does not contain any hardcoded credentials, obfuscated URLs, or remote code execution patterns. It depends on a well-known utility, 'treezip', for efficient tree scanning.
Audit Metadata