2026-coach
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses a role-play instruction in SKILL.md ('You are now acting as an executive coach helping the user plan their 2026') to define the agent's persona. While this is the primary purpose of the skill, role-reassignment patterns are monitored as potential injection vectors.
- [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted user input during the 'Discovery Phase' and interpolates that data into generated files (2026_PLAN.md, COACHING_CONTEXT.md).
- Ingestion points: User responses to four rounds of discovery questions defined in SKILL.md (e.g., current role, vision, strategy, daily behaviors).
- Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions in the user's input.
- Capability inventory: The skill possesses file-writing capabilities to save artifacts in the user's current directory or home directory (SKILL.md, Phase 5).
- Sanitization: The instructions do not define sanitization or escaping protocols for external content before it is written to the strategic plan files.
Audit Metadata