The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions define a rigid architectural planning workflow. No patterns attempting to bypass safety filters, extract system prompts, or override agent constraints were detected.
[DATA_EXFILTRATION]: The skill operates on project-specific documentation paths (e.g., docs/plans/, docs/editor-behavior/). It does not contain network exfiltration commands or access to sensitive user credentials or system files.
[REMOTE_CODE_EXECUTION]: The skill references other agent capabilities (lenses) like tdd, performance-oracle, and research-wiki to perform analysis. It does not download or execute arbitrary remote scripts.
[COMMAND_EXECUTION]: The workflow involves file system reads and writes to specific documentation directories. No unsafe subprocess spawning or arbitrary shell command execution patterns were found.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private secrets were found within the instructions or metadata.
[OBFUSCATION]: The content is provided in clear text with no evidence of Base64 encoding, zero-width characters, or other obfuscation techniques.
[INDIRECT_PROMPT_INJECTION]: While the skill reads external documentation and research, it implements a 'Confidence Score' system that requires cited evidence and verification for all claims, serving as a defensive layer against untrusted data influencing the plan's verdict.

plate-plan