The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes infrastructure configuration data from cloud provider APIs (AWS, Azure, GCP), which could be manipulated to contain malicious metadata. The skill possesses capabilities to write files and modify cloud resources based on this data.
Ingestion points: scripts/agent.py retrieves data via boto3 API calls (e.g., iam.get_account_summary, ec2.describe_vpcs); SKILL.md includes workflows that parse prowler and scoutsuite JSON output using jq.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to ignore potentially malicious instructions embedded within the retrieved cloud resource metadata.
Capability inventory: The Python agent script writes audit reports to cis_audit_report.json; the SKILL.md workflow includes shell commands for sensitive operations such as deleting IAM keys and modifying S3 bucket encryption.
Sanitization: The skill lacks explicit sanitization or validation routines for the cloud metadata before it is processed or used in remediation steps.
[COMMAND_EXECUTION]: The skill provides instructions and examples for executing shell commands to perform security audits and apply remediation fixes to cloud environments.
Evidence: SKILL.md contains several administrative commands, including aws iam delete-access-key, aws s3api put-bucket-encryption, and automated scanner invocations (prowler aws, scout suite aws).
[PERSISTENCE]: The skill suggests establishing automated, recurring audit tasks using system scheduling utilities.
Evidence: SKILL.md includes a crontab example (0 2 * * 1 prowler aws ...) to maintain periodic security assessment across sessions.

auditing-cloud-with-cis-benchmarks