The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions require the agent to execute shell commands such as hugo, ls, and find to verify build status and file structure of artifacts during quality gate checks (SKILL.md).
[DYNAMIC_EXECUTION]: The skill directs the agent to execute tools using absolute file paths that may be resolved by other skills (e.g., /opt/skills/storage-audit/scripts/storage-audit), which involves running local binaries outside of the standard PATH (SKILL.md, evals.json).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and verify untrusted external artifacts and claims. Instructions embedded in these artifacts could potentially influence the agent's verification process.
Ingestion points: Processes artifacts, implementations, and source documents specified in user requests (SKILL.md).
Boundary markers: The skill uses a structured verification protocol but lacks explicit delimiters or sanitization for the content of the artifacts being analyzed.
Capability inventory: Access to shell commands (ls, find, hugo, search_files, read_file), binary execution, and network requests (curl).
Sanitization: No explicit sanitization or filtering of artifact content is documented before the investigation phase.
[DATA_EXFILTRATION]: The skill utilizes curl to perform network requests against URLs found within potentially untrusted artifacts to verify link resolution (SKILL.md).

verification-methodology