The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to scan external project files (README, docs, code) to gather context for planning. This creates a surface for indirect prompt injection where malicious instructions inside those files could influence the generated plan.
Ingestion points: SKILL.md (Workflow section) specifies reading README.md, documentation, and code files.
Boundary markers: Absent.
Capability inventory: The skill is limited to generating a markdown checklist and does not perform file writes, command execution, or network operations.
Sanitization: Absent.
[SAFE]: The skill consists entirely of high-level natural language instructions for organizing tasks and does not include any executable code, scripts, or network access.

concise-planning