The Agent Skills Directory

[COMMAND_EXECUTION]: The short-alpha-battle-runner.py script uses subprocess.run to orchestrate a sequence of local Python scripts, including short-alpha-regime-engine.py, short-alpha-event-clock.py, and short-alpha-terrain-engine.py. This pattern manages the entire strategy generation workflow through local subprocess execution.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it processes untrusted external financial data to influence its decision-making logic.
Ingestion points: Untrusted data enters the agent context through sources/news-event-log.jsonl and results/candidates/candidates.json, which are read by short-alpha-market-observe.py.
Boundary markers: None; the skill does not use specific delimiters or instructions to prevent the agent from obeying instructions that might be embedded in news headlines.
Capability inventory: The skill can execute local scripts via subprocess.run and perform file system writes across the workspace (e.g., in short-alpha-strategy-register.py).
Sanitization: Numerical data is passed through to_float and clamp functions, but text-based headlines are evaluated directly for heuristic triggers (e.g., searching for keywords like 'war' or 'ceasefire').
[COMMAND_EXECUTION]: The short-alpha-battle-runner.py script modifies file permissions via dst.chmod(0o755) for its internal Python and Bash scripts when synchronizing them to a task directory to ensure they can be executed by the pipeline.
[COMMAND_EXECUTION]: Bash scripts such as short-alpha-general-bootstrap.sh and short-alpha-general-smoke.sh perform shell operations to initialize task directories, copy script assets, and automate the integration testing of the analytical pipeline.

short-alpha-general