The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts scripts/agent.py and scripts/process.py execute the cosign CLI using subprocess.run. The commands are invoked using argument lists, which is a secure implementation that prevents shell injection attacks.
[EXTERNAL_DOWNLOADS]: SKILL.md provides installation instructions that fetch the cosign binary from Sigstore's official GitHub releases. Sigstore is the primary vendor for the tool and is a well-known security project.
[INDIRECT_PROMPT_INJECTION]: The skill reads and parses JSON data returned by the cosign tool to generate reports and policies.
Ingestion points: cosign CLI standard output in scripts/agent.py and scripts/process.py.
Boundary markers: None identified; the skill relies on JSON structure for data delimitation.
Capability inventory: The skill has the ability to execute system commands via subprocess.run and write files to the local disk.
Sanitization: The data is parsed using the standard Python json.loads() method before use.
[PRIVILEGE_ESCALATION]: The SKILL.md documentation includes sudo commands (sudo mv and sudo chmod) for installing the cosign binary to a global system path. This is standard administrative practice for installing command-line tools.

implementing-image-provenance-verification-with-cosign