The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and parse WorkLog files from the project environment or user uploads to maintain session state. While this creates a surface for indirect prompt injection if external data is manipulated, the risk is mitigated by the skill's core workflow which requires a 'STOP' and 'Present WorkLog to user' step for human verification before proceeding. Ingestion points include WorkLog.md in the project root, uploaded files in /mnt/user-data/uploads/, and files in /mnt/project/. The use of YAML frontmatter delimiters helps scope the metadata parsing.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard shell commands (cat, ls, cp, find, sed) to manage text-based WorkLog files. These commands are used solely for environment detection, file creation, and status updates within the local working directory. Evidence of these scripts can be found in the Environment Detection section of SKILL.md and the environment-specific reference files (e.g., references/codecli-environment.md). These operations are restricted to relevant project files and do not include high-risk command execution or privilege escalation.

iterating