The Agent Skills Directory

[PROMPT_INJECTION]: The file skill-report.json contains a security_audit section that uses self-referential language to explicitly declare the skill as "safe" and characterize any potential security findings as "false positives." This is a known technique (Category 8e) used to target and mislead security analysis tools.
[PROMPT_INJECTION]:
Ingestion points: skill-report.json metadata.
Boundary markers: Absent.
Capability inventory: The skill utilizes shell commands (grep, git, gh, ugig, sh1pt) and filesystem operations (creating staging directories in /tmp/).
Sanitization: Not present for the ingested metadata file.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a variety of shell commands, including repository management via gh and git, secret scanning via grep, and publishing via several third-party CLI tools (ugig, sh1pt, clawhub). These commands are documented and aligned with the skill's stated purpose of publishing artifacts.

promote-skill