The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a significant surface for indirect prompt injection by processing external, untrusted legal documents and providing a mechanism to update its 'Playbook' (source of truth) based on that processing.
Ingestion points: The skill ingests inbound NDAs (confidentiality agreements) provided by the user for triage (file: SKILL.md).
Boundary markers: The instructions do not define clear boundary markers or 'ignore' directives for the NDA content during processing.
Capability inventory: The agent has the capability to write to the local filesystem, specifically modifying the global configuration file at ~/.claude/plugins/config/claude-for-legal/commercial-legal/CLAUDE.md and writing matter-specific files to ~/.claude/plugins/config/claude-for-legal/commercial-legal/matters/.
Sanitization: There is no evidence of sanitization or validation of the ingested NDA text before it influences the modification of the core playbook file.
[COMMAND_EXECUTION]: The skill directs the agent to perform file system write operations to manage legal matters and update configuration settings.
Evidence: The skill instructs: 'Write outputs to the matter folder at ~/.claude/plugins/config/claude-for-legal/commercial-legal/matters//' and 'record the answer in ~/.claude/plugins/config/claude-for-legal/commercial-legal/CLAUDE.md'.

nda-review