The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external, potentially untrusted finding data and rules of engagement to generate a markdown report. While this creates a theoretical surface for indirect prompt injection (e.g., a malicious finding title appearing in the summary), the risk is nominal as the output is intended for human review and does not feed back into automated execution paths.
Ingestion points: all-with-owasp.jsonl, owasp-coverage.md, and roe.yaml (read by scripts/exec_summary.py).
Boundary markers: No explicit delimiters or instructions are used to separate ingested data from the report template in the output markdown.
Capability inventory: The skill is restricted to reading and writing local files and executing Python scripts; network access and dangerous bash commands (e.g., rm, curl, wget) are explicitly disallowed in the frontmatter.
Sanitization: Data is parsed via standard JSON and YAML libraries before being interpolated into a static markdown template using Python f-strings.
[SAFE]: The skill implements security best practices by defining a clear disallowed-tools list, preventing the use of networking tools or deletion commands. All code is contained within the local environment, and external dependencies are limited to standard libraries.

generating-executive-summary