The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands via the git CLI to manage the environment for its assessment. This includes switching branches (git switch), fetching updates (git pull), and creating worktrees (git worktree add). These operations are consistent with the skill's primary purpose of analyzing different versions of a codebase.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it reads and processes untrusted data from the repository being analyzed.
Ingestion points: The skill reads project documentation (README.md, PRD.md, docs/architecture/), source code files (e.g., src/domain/orden.ts), and git diff output.
Boundary markers: No explicit delimiters or boundary markers are defined to isolate untrusted data from the agent's instructions during the evaluation phase.
Capability inventory: The agent can execute git commands for branch/worktree management and write assessment files to the docs/assessments/ directory.
Sanitization: There is no evidence of sanitization or filtering of the content read from the repository before it is processed by the AI for architectural evaluation.

fmway-arch