The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/scan.sh and scripts/scan-all.sh files execute shell commands such as ripgrep (rg), grep, and find to perform filesystem-wide searches for sensitive patterns.
[DYNAMIC_EXECUTION]: The scripts/scan-all.sh script dynamically executes external scripts located in the user's home directory (~/.config/opencode/skill/) based on filesystem presence checks. It also utilizes eval to process conditional logic strings to determine which security scans should be activated.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process arbitrary data from external repositories during its scanning process. While SKILL.md provides instructions for the agent to redact findings, the underlying shell scripts output raw matches to standard output, which could contain malicious payloads designed to influence the agent's behavior.
Ingestion points: File content retrieved by rg or grep in scripts/scan.sh from the target scanning directory.
Boundary markers: No programmatic boundary markers are present in the execution scripts; the skill relies on instructional constraints in SKILL.md to guide agent behavior.
Capability inventory: The skill possesses the ability to execute shell commands and invoke external sub-scripts.
Sanitization: No programmatic sanitization or escaping of scan results is implemented in the scripts; raw file matches are printed directly to the console.

security-secrets