The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's TASKSOLVER role is designed to read project registers (e.g., TODO.md, AUFGABEN.txt) and autonomously solve "trivial" tasks. This creates a surface where untrusted data within these files can influence the agent's behavior.
Ingestion points: Reads arbitrary content from project-specific files like TODO.md, README.md, and CHANGELOG.md (SKILL.md, Role 1 & 2).
Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded prompts within the files it processes.
Capability inventory: The skill has the ability to write to the file system (updating documentation), move files (archiving strays), execute internal CLI tools (usmc), and orchestrate other protocols like bilingual-doc-sync.
Sanitization: No sanitization or validation of the content read from the project files is specified before the agent acts upon it.
[COMMAND_EXECUTION]: The skill uses a command-line utility usmc to log session metadata and project status.
Evidence: usmc --agent <agent> note "tidy-up: project=<project-path-or-slug> at=<ISO-time> result=<done|nothing-to-do>" --type context --priority 2 --tags "tidy-up-log,<project-slug>" (SKILL.md, Due-criterion section). This appears to be a vendor-specific management tool and is used within its intended scope.

tidy-up