The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to aggregate data from multiple local JSON and YAML files located in the reports/ directory. Because these files are external inputs, they represent an attack surface where maliciously crafted signal data could manipulate the aggregate conviction scores or mislead an agent reviewing the resulting dashboard.
Ingestion points: The script scripts/aggregate_signals.py ingests data through load_json_files and load_yaml_files from user-defined paths targeting multiple upstream skill outputs.
Boundary markers: There are no specific boundary markers or instructions to the agent to ignore potential injections within the raw signal titles, descriptions, or ticker lists during the aggregation process.
Capability inventory: The skill is capable of reading files from the filesystem and writing JSON/Markdown reports to the reports/ directory. It does not possess network access, shell execution, or dynamic code evaluation capabilities.
Sanitization: The skill performs strict type checking, uses yaml.safe_load for configuration, and implements a comprehensive parse_raw_score function to validate and normalize numeric inputs, which mitigates technical data corruption but not semantic injection.

edge-signal-aggregator