The Agent Skills Directory

[COMMAND_EXECUTION]: The included audit script scripts/audit-pack.py executes git via subprocess.run to extract metadata such as the repository origin and commit hash. This execution is limited to informational queries used to verify the provenance of the audited files.- [EXTERNAL_DOWNLOADS]: The skill instructions utilize npx to download and run the skills CLI package at a specific version (1.5.23). This is a standard tool for skill lifecycle management on this platform.- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and audit untrusted third-party code from the bam-bam-2/solo-skills repository, which creates an indirect prompt injection surface.
Ingestion points: The scripts/audit-pack.py script and the agent's inspection instructions involve reading content from all files within the external repository.
Boundary markers: The skill contains comprehensive safety documentation (references/external-actions-and-safety.md) that explicitly warns the agent to identify, flag, and ignore embedded instructions or bypass flags within the ingested data.
Capability inventory: The skill is configured with file manipulation tools (Read, Write, Edit) and shell execution capabilities (via Bash and the Python helper) to perform the audit and adaptation.
Sanitization: The 'Portability worksheet' and adaptation instructions require the agent to strip all author-specific assumptions, credentials, and permission-bypass flags from the third-party content before installation.

solo-skills