The Agent Skills Directory

[DYNAMIC_EXECUTION]: The Python scripts scorecard.py and dashboard.py dynamically modify the Python search path (sys.path) to include relative directories. Specifically, scorecard.py attempts to import a dependency from a sibling fyers-trading directory. This pattern is used for local skill interdependency but involves loading code from computed paths, which can be risky if the local file system structure is manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill operates by analyzing trading strategy definitions and backtest evidence to propose improvements. Because the agents process and act on natural language content within these external strategy files, there is a risk that malicious instructions could be embedded to influence the optimization process.
Ingestion points: Strategy definitions and backtest data are read and analyzed in Phase 1 (Understand) of the optimization loop as described in SKILL.md and references/optimization-loop.md.
Boundary markers: The skill does not implement specific delimiters or instructions to ensure that natural language content within the strategy files is not interpreted as directives by the council agents.
Capability inventory: The skill utilizes powerful tools including Bash for executing backtests, Write/Edit for code generation, and the Agent tool for orchestrating sub-processes.
Sanitization: No sanitization or filtering is specified for the natural language content extracted from strategy files before it is processed by the AI agents.
[COMMAND_EXECUTION]: The skill directs the host agent to use the Bash tool to run backtests on strategy variants. This involves the execution of code generated by the AI (the Variant Generation Strategist). While necessary for the skill's primary purpose, this recursive generation and execution loop necessitates user oversight to ensure that optimized variants do not contain unintended or malicious shell commands.

fyers-supercharge