The Agent Skills Directory

[COMMAND_EXECUTION]: The evaluation script evals/run-static-checks.sh dynamically locates and executes a library file (static-check-lib.sh) by traversing parent directories at runtime using the source command.
[COMMAND_EXECUTION]: The core instructions in SKILL.md require the agent to execute just check (a shell-based task runner) after directly editing files, providing a path for arbitrary command execution if the edits are influenced by malicious input.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from other skill files while possessing significant system capabilities.
Ingestion points: The agent is instructed to read SKILL.md, skill.json, references, and evals for any target skill in the catalog (skills/*/, .agents/skills/*/).
Boundary markers: Absent; there are no instructions or delimiters designed to prevent the agent from following malicious commands embedded in the files being reviewed.
Capability inventory: The agent has authority to edit files directly and execute shell commands via the just utility.
Sanitization: Absent; the skill does not specify any sanitization, filtering, or validation of the content read from external skill files before use in the review process.

skill-reviewer