The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands to interact with the local environment and fetch data.
Evidence: scripts/review.py executes openclaw senpi events to retrieve runtime telemetry logs and curl to query external APIs.
[EXTERNAL_DOWNLOADS]: The skill fetches trading information from a well-known external service.
Evidence: scripts/review.py makes network requests to https://api.hyperliquid.xyz/info using curl to recover fill data for closed trading strategies.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external APIs and local logs which could be manipulated to influence the agent's behavior.
Ingestion points: Data is retrieved from the Hyperliquid API, local telemetry event logs via openclaw, and Senpi MCP tools such as discovery_get_trader_history.
Boundary markers: The skill instructions in SKILL.md define nine specific 'guardrails' and 'Hard Rules' that explicitly instruct the agent to use only engine-computed values, lead with total PnL, and avoid following potentially biased external directives.
Capability inventory: The skill can read local strategy mandates, perform network requests for data retrieval, and execute specific command-line tools.
Sanitization: External data is parsed as JSON and processed through deterministic logic in scripts/review.py (e.g., numeric normalization) before being passed to the agent.
[SAFE]: The skill implements a custom, restricted YAML loader in scripts/_yaml.py to prevent unsafe deserialization risks (like those associated with standard Python YAML loaders).

senpi-improve-trades