The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard system utilities such as uname, git, and claude, as well as vendor-specific scripts like loci_stats.py. These commands are used solely for gathering environment state and diagnostic information.
[DATA_EXFILTRATION]: Collects project configuration details (e.g., project-context.json, hooks.json) and system metadata to generate a local markdown report. No data is transmitted to external servers or non-whitelisted domains.
[SAFE]: An indirect prompt injection surface exists because the skill incorporates untrusted data from project files and user descriptions into its report.
Ingestion points: Reads local configuration JSON files and the user's problem description string.
Boundary markers: The report uses markdown structural elements and <details> tags to isolate data.
Capability inventory: Uses Bash and Read tools for diagnostic purposes; no high-risk capabilities like network writing are present.
Sanitization: File contents are included in the report verbatim without specific sanitization filters.

bug-report