The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/pipeline/strategy_exporter.py dynamically loads and executes a Python module from a calculated relative path (../../../../edge-candidate-agent/scripts/candidate_contract.py). The function _load_candidate_contract_validator uses importlib.util.spec_from_file_location and spec.loader.exec_module(module) to run code from a location outside the skill's own folder structure, which is a significant security risk if the environment's directory structure is manipulated.
[INDIRECT_PROMPT_INJECTION]: The skill takes an external JSON input bundle and processes fields like journal_snippets and observations which are directly interpolated into the prompt templates. Specifically, scripts/pipeline/evidence_extractor.py formats this data, and it is then injected into prompts/developer_prompt_template.md using the {{evidence_summary}} tag.
Ingestion points: The --input JSON bundle parsed in scripts/run_hypothesis_ideator.py.
Boundary markers: The prompt uses Markdown headers to separate sections, but does not provide specific delimiters or instructions to ignore embedded commands within the user-provided data.
Capability inventory: The skill has the capability to write various JSON, YAML, and MD files to the output directory and execute local code via dynamic loading.
Sanitization: The _as_list_of_str function ensures data is in string format and stripped of whitespace, but no active filtering or sanitization against prompt injection techniques is performed.

trade-hypothesis-ideator