The Agent Skills Directory

[SAFE]: The skill is designed with a strong defensive posture, emphasizing safety gates, referral protocols, and clear scope boundaries for coaching interactions.
[INDIRECT_PROMPT_INJECTION]: The skill maintains an attack surface for indirect prompt injection as it processes user-provided coaching goals and reflective inputs.
Ingestion points: User prompts requesting coaching, ambivalence exploration, or habit design (documented in SKILL.md and README.md).
Boundary markers: Mandatory session gates requiring the agent to read safety and scope references before responding, and conversational micro-agreements to define session limits.
Capability inventory: A local manifest validator script (scripts/validate-capabilities.py) and worksheet templates.
Sanitization: The validation script enforces strict schema and value constraints on host-owned configuration manifests to prevent unauthorized capability activation.
[COMMAND_EXECUTION]: The bundled test suite (tests/test_validate_capabilities.py) utilizes subprocess.run to verify the command-line behavior of the skill's internal validator script. This is restricted to verifying local file processing and does not involve network operations or untrusted input execution.

life-coach