sota-observability
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to perform end-to-end audits of existing codebases and deployments, which involves ingesting and analyzing untrusted external content such as source code, configuration files, and log data.
- Ingestion points: The 'AUDIT mode' described in
SKILL.mdandrules/06-audit-playbook.mdexplicitly instructs the agent to 'sample real code paths' and read configuration manifests across various files such asrules/06 §1andrules/05 §7a(edge access logs). - Boundary markers: While the instructions emphasize 'do not trust README claims' (
rules/06 §1), there are no explicit delimiters or 'ignore embedded instructions' warnings for the content being audited to prevent accidental obedience to instructions hidden in data. - Capability inventory: The skill leverages
grepand file-reading capabilities to scan and evaluate the observability posture of the target system across various files (e.g.,rules/01,rules/05). It also suggests analyzing live telemetry streams and alert histories (rules/06 §1). - Sanitization: The skill lacks formal sanitization, validation, or escaping mechanisms for the external content it processes, relying on the agent's ability to parse code and configuration without executing embedded natural language instructions.
Audit Metadata