The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses several shell commands to manage the auditing process, including git ls-files, find, wc, git rev-parse, git checkout, and git diff. These tools are used for project scope enumeration, cost estimation, and the application and verification of code fixes, granting the agent broad execution capabilities over the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill's design is susceptible to indirect prompt injection attacks because it reads and processes untrusted data from the repository without sufficient safeguards.
Ingestion points: Repository files are ingested for classification in Phase 1 (guide-phases-0-2-consent-scope-health.md) and for deep logic tracing in Phase 3 (guide-phases-3-5-review-locate-clarify.md).
Boundary markers: Absent. The skill instructions do not specify the use of delimiters or 'ignore' instructions to prevent the agent from executing commands hidden within the source code or documentation it reviews.
Capability inventory: The skill possesses the capability to modify any file in the repository during the 'Apply + Verify' stage in Phase 7 (guide-phases-6-9-fix-iterate-report.md).
Sanitization: Absent. There is no mention of escaping, validation, or filtering of repository content before it is evaluated by the logic-review methodology.
[METADATA_POISONING]: A discrepancy exists between the reported skill author ('sickn33') and the source repository metadata ('hyhmrright/logic-lens'). This mismatch in attribution could lead to confusion regarding the skill's origin and the trustworthiness of its automated logic.

logic-fix-all