The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions direct the agent to resolve and read a private Ed25519 signing key from the filesystem. Specifically, it searches for a JSON file containing a 64-character hex private key at verification/witness-key.json or a path specified by the RUFLO_WITNESS_KEY_PATH environment variable.
[EXTERNAL_DOWNLOADS]: The skill uses npm install to download and install the neural-trader package from the public npm registry at runtime if it is not already present. This introduces a dependency on untrusted external code.
[COMMAND_EXECUTION]: The skill performs arbitrary command execution using Bash and npx to run backtests, manage npm packages, and execute local scripts located at plugins/ruflo-neural-trader/src/signed-artifact.mjs.
[DATA_EXPOSURE]: The skill accesses and processes sensitive financial strategy configurations and performance metrics from the memory_store and agentdb_pattern-store tools.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection. It ingests untrusted data from backtest output metrics and strategy configurations stored in memory. It lacks explicit boundary markers or sanitization logic when interpolating these results into further processing steps, while maintaining high-privilege capabilities like command execution and network-adjacent operations (via npm).

trader-backtest