The Agent Skills Directory

[DATA_EXFILTRATION]: The instruction to read "EVERY SOURCE FILE IN FULL" creates a significant risk of data exposure, as it encompasses sensitive files such as environment variables (.env), credentials, and private project configurations that are typically excluded from general context processing.
[PROMPT_INJECTION]: The skill uses authoritative language ("critical and non negotiable") designed to override the agent's internal constraints regarding resource limits, token usage, or safety-based selective reading.
[PROMPT_INJECTION]: The skill enables indirect prompt injection by systematically ingesting all codebase content, creating a large surface area for malicious instructions embedded in untrusted source files to execute within the agent's context.
Ingestion points: The agent is instructed to use the Read tool on every file in the codebase.
Boundary markers: Absent; the skill lacks delimiters or instructions to ignore commands found within the processed files.
Capability inventory: The skill explicitly directs the use of the Read tool with paging capabilities (offset, limit).
Sanitization: Absent; no filtering or validation is performed on the ingested file contents.
[COMMAND_EXECUTION]: The skill provides explicit logic for using external tools (Read) to systematically traverse the filesystem and retrieve full file contents.

learn-codebase