The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the conviso CLI tool to perform management tasks on the Conviso Platform. Additionally, several shell scripts (10_collect_recent_vulns.sh, 20_prioritize_vulns.sh, 30_generate_bulk_update_csv.sh, 35_validate_bulk_csv.sh) use python3 to execute logic provided via heredocs. This pattern involves the runtime execution of dynamically provided script content.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes vulnerability data (titles and descriptions) retrieved from the Conviso API.
Ingestion points: External vulnerability data is fetched and stored in out/recent_vulns.json by scripts/10_collect_recent_vulns.sh.
Boundary markers: The SKILL.md file defines safety rules for the agent, and the scripts/20_prioritize_vulns.sh script includes a safe_text function to strip markdown headers and control characters.
Capability inventory: The skill has the ability to perform bulk updates to the Conviso Platform via scripts/50_bulk_apply.sh.
Sanitization: Input is sanitized by stripping backticks and hashes and truncating strings to 180 characters, which provides some protection against accidental instruction obedience.

conviso-vuln-remediator