The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file defines a 'Decision feedback protocol' which instructs the agent to execute a local Python script utilities/skill-builder/scripts/record_skill_feedback.py. This script is used to persist session metadata such as decisions and outcomes using shell command parameters.
[PROMPT_INJECTION]: The skill processes external, potentially untrusted data such as 'user request', 'PRD', and 'ticket' as listed in the 'Required inputs' section of SKILL.md. It lacks explicit boundary markers or sanitization procedures to isolate these inputs from the core instructions, creating a surface for indirect prompt injection.
Ingestion points: SKILL.md (user request, draft idea, PRD, ticket, screenshots)
Boundary markers: Absent
Capability inventory: SKILL.md (local script execution via python3)
Sanitization: Absent (only general instructions to redact secrets are provided)

brainstorming