The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agent.py utilizes the subprocess.run function to execute various cloud CLI tools (AWS, Azure, and Google Cloud). While this is the primary intended function of the skill for performing security operations, it grants the agent the capability to run system-level commands.
[METADATA_POISONING]: An inconsistency exists between the skill's metadata and its legal documentation; the SKILL.md frontmatter lists 'mahipal' as the author, whereas the LICENSE file attributes copyright to 'mukul975'.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically AWS CloudTrail audit logs, which contain attacker-influenced fields such as User-Agents and resource names.
Ingestion points: The query_cloudtrail_for_key function in scripts/agent.py fetches raw event data from AWS CloudTrail.
Boundary markers: None detected. Data is parsed directly into reports without delimiters or instructions for the agent to ignore embedded commands.
Capability inventory: The skill has the ability to write IAM policies, deactivate access keys, and execute arbitrary cloud CLI commands via subprocess.run.
Sanitization: Minimal sanitization is performed; while the userAgent field is truncated to 100 characters, other fields are processed without escaping or validation.

detecting-compromised-cloud-credentials