The Agent Skills Directory

[PROMPT_INJECTION]: Documentation in 'references/PHASES.md' contains various examples of prompt injection techniques (e.g., 'ignore previous instructions', 'DAN' jailbreak, role reassignment) for educational and detection-reference purposes. These do not attempt to override the agent's current system instructions or behavior.
[EXTERNAL_DOWNLOADS]: The 'scripts/setup.sh' file downloads the Sigil CLI binary from the author's GitHub repository ('github.com/NOMARJ/sigil'). This is an expected vendor resource and the script includes SHA256 checksum verification to ensure file integrity.
[DATA_EXPOSURE]: The 'scripts/audit-env.sh' script audits sensitive local files, including SSH private keys ('~~/.ssh/id_rsa'), AWS credentials ('~~/.aws/credentials'), and shell history files. This access is consistent with the skill's stated purpose of environment auditing; results are reported locally to the user and no network exfiltration was detected.
[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves scanning external repositories, third-party skills, and remote URLs. This creates an inherent surface for indirect prompt injection where malicious code in a scanned target could attempt to influence the agent's subsequent analysis or presentation of results.
[DYNAMIC_EXECUTION]: The skill uses Python one-liners within shell scripts ('scripts/_lib.sh', 'scripts/audit-skills.sh') to process JSON data and executes the downloaded binary ('sigil') as part of its primary scanning workflow.

sigil-scan