The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses clear, instructional language focused on legal template generation. No patterns associated with bypassing safety filters, developer mode overrides, or DAN-style injections were detected.
[DATA_EXPOSURE]: The skill scans project files like docs/prd.md and CLAUDE.md to identify data collection practices for disclosure in legal documents. It does not access sensitive credential files, environment variables, or private keys.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns or attempts to download and execute scripts from external sources were found. The skill relies on local templates and project analysis.
[COMMAND_EXECUTION]: The skill uses platform-provided tools (Read, Grep, Write) for file management. There are no attempts to execute arbitrary shell commands or system-level operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local data from files like docs/prd.md to populate templates.
Ingestion points: Reads docs/prd.md, CLAUDE.md, and templates/stacks/{stack}.yaml.
Boundary markers: None explicitly defined, though the instructions guide the agent to specific data extraction tasks.
Capability inventory: Limited to reading, writing Markdown files, and asking the user questions.
Sanitization: None specified, however, the output is restricted to static legal templates for user review, which significantly mitigates the risk of malicious payload execution.
[OBFUSCATION]: The content is written in clear Markdown and YAML. No encoded strings, homoglyphs, or zero-width characters intended to hide malicious behavior were identified.
[PERSISTENCE]: The skill does not attempt to modify startup scripts, scheduled tasks, or system configurations to maintain access.

solo-legal