The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution using tools like git, mkdir, chmod, install, and gh to profile repositories, manage temporary files, and interact with issue trackers.
[DYNAMIC_EXECUTION]: Employs large blocks of embedded Python code within shell wrappers (ultima.sh, tickets.sh) to handle complex logic for repository profiling, result merging, and rendering the final HTML report.
[INDIRECT_PROMPT_INJECTION]: As an auditing tool, the skill is designed to process untrusted codebase content. While it includes robust instructions to treat this content as data and mandates the redaction of secrets, its capabilities—specifically filing tickets and applying local code fixes—represent a potential attack surface.
Ingestion points: Source code, manifests, and documentation read from the repository being audited during profiling and lens dispatch.
Boundary markers: SKILL.md contains explicit instructions to treat tool output and instructions in untrusted documents as data. Subagent prompts are structured with XML-like tags to maintain separation.
Capability inventory: Includes scripts/tickets.sh for external ticket creation, references/fix-one.md for local file modification, and various read/write operations for reporting.
Sanitization: Mandatory instructions to redact sensitive and personal data from all outputs using the [REDACTED] placeholder.

ultima