affirmations
Warn
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation includes a command
npx skills add inference-sh/skills@prompt-engineeringin the 'Related Skills' section. The source 'inference-sh' is not a verified trusted vendor or a well-known service. Suggesting the installation and execution of external packages from unverified sources poses a supply chain risk. - [PROMPT_INJECTION]: The skill uses a set of behavioral triggers (e.g., 'stuck', 'looping', 'thrashing', 'I keep failing') in its metadata that instruct the agent to change its current execution trajectory and adopt a new set of rules.
- Ingestion points: The triggers are defined in the YAML frontmatter 'description' field of SKILL.md and are intended to be activated by the agent's self-observation or user feedback.
- Boundary markers: There are no explicit boundary markers or instructions to ignore these triggers if they appear in untrusted external data.
- Capability inventory: The skill itself is instructional and does not define its own tools, though it influences the use of existing agent capabilities.
- Sanitization: No sanitization is present to prevent indirect injection of these triggers from external data sources.
Audit Metadata