The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The 'SELF-EVOLUTION TELEMETRY' section instructs the agent to search for specific project paths within the hidden directory ~/.claude/projects/. While intended for logging, this involves accessing sensitive platform-specific state folders in the user's home directory.
[PERSISTENCE]: The skill attempts to maintain a log of its executions by writing to skill-telemetry.md within the ~/.claude/projects/ directory. Automated writing to hidden system/platform directories without explicit user consent is a persistence risk.
[PROMPT_INJECTION]: The instructions include autonomy-forcing directives such as 'Do NOT ask questions' and 'Investigate the entire codebase systematically', which aim to override the agent's standard conversational and safety-checking behaviors.
[INDIRECT_PROMPT_INJECTION]: The skill is highly vulnerable to indirect prompt injection due to its core function of processing untrusted external data.
Ingestion points: The skill reads the 'entire grid optimization codebase' from the current working directory and accepts user-provided $ARGUMENTS.
Boundary markers: Absent. The instructions do not specify any delimiters (e.g., XML tags or triple backticks) to separate the analyzed codebase content from the agent's instructions.
Capability inventory: The skill possesses extensive analysis capabilities and performs file system read operations across various tech stacks. It also includes the ability to write telemetry data to the home directory.
Sanitization: Absent. There is no requirement or logic provided to sanitize, escape, or filter instructions that might be embedded in the code files being analyzed.

grid-optimizer