fatigue-frequency-manager
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown and schema files. There is no executable code (Python, JavaScript, or shell scripts) included in the skill, which eliminates the risk of malicious code execution or system-level compromise.
- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to ingest external data (scraped text, reviews, and project files), it includes explicit instructions to treat this data as untrusted input. The ingestion points are user-provided context and connected sources; boundary markers are present in the 'Data Sources' and 'Unified Skill Contract' sections (instructing the agent to treat material as evidence, not instructions); capability inventory shows no internal execution tools; and sanitization is addressed through agent instructions to label evidence types and distinguish proxies from observations.
- [DATA_EXFILTRATION]: The skill documentation establishes clear security protocols, specifically stating that API keys must be kept out of skills and source control in 'references/connectors.md', and requiring explicit user authorization for any persistent writes or external side effects.
- [METADATA_POISONING]: The YAML frontmatter and metadata fields are consistent with the skill's stated purpose of ad fatigue management and show no signs of deceptive behavior or malicious redirection.
Audit Metadata