The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from a repository's source files. An attacker could embed malicious instructions within code comments or files that are subsequently read or grepped by the agent.
Ingestion points: Source files searched and read via grep or read operations.
Boundary markers: Absent; the skill provides no instructions to ignore embedded directives in the search results.
Capability inventory: The skill leverages grep and file-reading capabilities which could be exploited if an injection succeeds.
Sanitization: None; the content from the codebase is processed as-is.
[DATA_EXFILTRATION]: The skill's primary purpose includes locating sensitive artifacts such as environment variables (env vars) and specifically mentions searching configuration-related files like .env.sample and setTestEnv.sh. This could be used to discover sensitive configuration structures or secrets.
[COMMAND_EXECUTION]: The workflow relies on the execution of shell commands like grep to perform broad discovery across the codebase.

source-search