alspec
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses imperative language to override the agent's standard behavior. Specifically, it uses the phrase "Do not use your own plan mode" to force the agent to abandon its default operational framework in favor of an external script.
- [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by instructing the agent to read and execute content from files outside its immediate scope.
- Ingestion points: The agent is directed to read
../alignfirst/SKILL.mdandreferences/spec-protocol.mdat runtime. - Boundary markers: No delimiters or instructions to ignore embedded commands within the external files are provided.
- Capability inventory: No direct system or network capabilities are defined in this file, but the skill delegates execution logic to the external files.
- Sanitization: There is no evidence of sanitization or validation of the content retrieved from the referenced external files.
Audit Metadata