The Agent Skills Directory

[SAFE]: The skill is strictly read-only and explicitly prohibits writing to specification files or source code, which minimizes the risk of unauthorized modifications to the project environment.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it processes untrusted data from external files.
Ingestion points: The agent reads SPEC.md and various source files to perform drift detection as directed in SKILL.md.
Boundary markers: Absent; the skill does not define specific delimiters or instructions to ignore embedded commands that might be found in the specification or code files.
Capability inventory: Capabilities are limited to grep and file reading within the local filesystem; no network access or file-writing tools are utilized (SKILL.md).
Sanitization: Absent; no explicit sanitization or filtering of the ingested content is performed before processing.
[SAFE]: No external network operations, remote code downloads, or exfiltration patterns were identified.
[SAFE]: The skill does not request or use administrative privileges and does not include any persistence mechanisms.

check