The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates Markdown reports by reading finding data from .sb-pentest-context.json. As this data represents the results of security audits on external systems, it must be treated as untrusted. The skill provides no mechanisms to prevent the agent from executing instructions that might be embedded within these findings.
Ingestion points: The findings array within the .sb-pentest-context.json file.
Boundary markers: None; the skill does not instruct the agent to use delimiters or ignore instructions within the processed data.
Capability inventory: File system write access to create supabase-audit-report.md, write to .sb-pentest-audit.log, and update .sb-pentest-context.json.
Sanitization: There are no instructions for the agent to sanitize, escape, or validate the content of the findings before interpolating them into the final report.
[PROMPT_INJECTION]: The skill uses high-pressure, authoritative language to override the agent's standard operational workflow, mandating "progressive" file updates through imperative directives.
Evidence: Phrases such as "🔴 CRITICAL: PROGRESSIVE FILE UPDATES REQUIRED", "This is not optional. Failure to write progressively is a critical error.", and "FAILURE TO UPDATE CONTEXT FILES IS NOT ACCEPTABLE."

supabase-report