The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local bash scripts (get-memory-dir.sh) and a TypeScript validation script (schema.ts) to manage the project's .memory directory and ensure data consistency. These operations are performed within the local environment and are intended for project bookkeeping.
[PROMPT_INJECTION]: The skill incorporates an indirect prompt injection surface because it reads and updates markdown files in the .memory/ directory which are used to track tasks, research, and project knowledge.
Ingestion points: Files like .memory/todo.md, .memory/summary.md, and .memory/research-*.md are read as source of truth for the project state.
Boundary markers: The skill does not use explicit boundary markers or 'ignore' instructions when reading the body of these markdown files.
Capability inventory: The agent can execute shell scripts, read/write files, and perform git operations.
Sanitization: The schema.ts script performs validation and repair of YAML frontmatter, but does not sanitize the markdown content itself. The risk is significantly reduced by the instruction that 'NEVER PUSH CHANGES WITHOUT HUMAN REVIEW'.

miniproject