The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides and generates numerous high-privilege CLI commands for AWS, Azure, and GCP. These include revoking IAM access keys, modifying network security groups, and creating disk snapshots. These operations are sensitive but consistent with the skill's stated purpose of incident response.
[INDIRECT_PROMPT_INJECTION]: The helper script scripts/process.py features an injection surface where user-provided resource details (from a JSON file) are interpolated directly into shell command strings without sanitization.
Ingestion points: The resource_json input file in scripts/process.py provides values for usernames, instance IDs, and bucket names.
Boundary markers: None are present to delimit user input or prevent shell meta-character execution.
Capability inventory: The generated commands possess capabilities to disable IAM credentials, modify network isolation rules, and manipulate cloud resources across three major providers.
Sanitization: Absent; the script uses f-strings for shell command construction, which may allow for command injection if the input data is sourced from an untrusted party.
[METADATA_POISONING]: There is a minor metadata discrepancy between the author name 'mahipal' specified in SKILL.md and the copyright holder 'mukul975' mentioned in the LICENSE file. While typically benign, such inconsistencies are noted for traceability.

performing-cloud-incident-containment-procedures