The Agent Skills Directory

[PROMPT_INJECTION]: The skill functions by ingesting and analyzing untrusted codebase content to identify regulatory compliance evidence. While this creates a surface for indirect prompt injection (e.g., through malicious code comments), it is the primary intended purpose of the skill. No specific boundary markers or instructions to ignore embedded commands are present in the instructions.
Ingestion points: Target codebase files, dependency manifests (package.json, requirements.txt, etc.), and user-provided $ARGUMENTS.
Boundary markers: Absent.
Capability inventory: Filesystem read access (entire codebase); local filesystem write access (telemetry logging).
Sanitization: Absent.
[COMMAND_EXECUTION]: The skill performs extensive filesystem operations to detect technology stacks and compliance evidence (e.g., reading configuration files, version control logs, and source code) as part of its auditing functions.
[DATA_EXFILTRATION]: The skill includes a self-evolution telemetry mechanism that appends execution metadata to a local file at ~/.claude/projects/skill-telemetry.md. This activity is restricted to the local filesystem and does not involve external network transmission.

manufacturing-compliance