The Agent Skills Directory

[SAFE]: The skill implements a robust task-management system using local project files to store state. All scripts provided are transparent and perform standard project management tasks.
[COMMAND_EXECUTION]: The skill utilizes a Python scaffolding script (scaffold.py) to generate project files and an initialization script (init.sh) to report project status. These operations are local, use standard system tools (git, python3), and are essential for the skill's functionality.
[DATA_EXPOSURE]: The skill reads project metadata, git history, and dependency status to populate its context. It does not attempt to access sensitive system files, environment variables, or credentials.
[INDIRECT_PROMPT_INJECTION]: The skill design relies on reading project-specific files like feature_list.json and progress.txt to determine future actions. This is a deliberate design choice for long-running agents to maintain state across sessions.
Ingestion points: feature_list.json, progress.txt, task.json are read from the local project directory.
Boundary markers: Absent; the agent is instructed to treat the content of these files as authoritative project state.
Capability inventory: The skill can write files and execute shell commands (via the scaffolding and initialization scripts).
Sanitization: Standard JSON parsing is employed; no explicit sanitization of descriptions or steps is performed.

task-harness