The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted codebase data, creating an attack surface for indirect prompt injection. * Ingestion points: The agent is instructed to read various files across the codebase during the 'PHASE 1: ECONOMY DISCOVERY' and 'PHASE 3: LOOT TABLE ANALYSIS' phases. * Boundary markers: There are no instructions to utilize delimiters or to disregard instructions found within the analyzed files. * Capability inventory: The skill utilizes file system read capabilities and directory traversal to evaluate economic logic. * Sanitization: There is no evidence of output sanitization or validation of the data processed from the codebase before it is included in the final analysis report.
[PERSISTENCE]: The skill implements a 'SELF-EVOLUTION TELEMETRY' mechanism that survives across different execution sessions. * Evidence: The instructions direct the agent to check for the existence of the ~/.claude/projects/ path and to 'append to skill-telemetry.md' if found. * Impact: This behavior allows the skill to maintain a persistent log of its activity and performance outcomes on the local machine.
[DATA_EXFILTRATION]: The skill performs sensitive file path access by interacting with platform-specific hidden configuration directories. * Evidence: The agent is instructed to explicitly search for and write to the ~/.claude/projects/ directory. * Impact: Accessing and writing to hidden directories (dot-folders) in the user's home directory is a sensitive operation that can expose project metadata and execution history.

game-economy