The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from repository files, including architecture records, manifests, and source code, to identify simplification candidates. This data ingestion is paired with high-privilege capabilities such as file system modification and command execution. Ingestion points include comprehensive repository scanning as described in SKILL.md and investigation.md. There are no explicit boundary markers or instructions to ignore embedded commands within the processed files. The skill has the capability to delete files and refactor code, and it lacks sanitization or validation for the ingested repository content.
[COMMAND_EXECUTION]: The instructions in execution-and-recovery.md direct the agent to discover and run repository-native verification commands like test, lint, and build suites. This results in the execution of arbitrary scripts defined within the target repository, which could be exploited if the codebase contains malicious automation scripts.

simplify-codebase