The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/reporting_gate.py to perform transparency linting. This script is intended to run via uv run python and is limited to using the Python standard library. The script performs deterministic checks on JSON data and does not have capabilities for network access, file system modification, or execution of arbitrary shell commands.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data in the form of a JSON manifest (claim.json) which is processed by the reporting gate script. This represents a potential surface for indirect prompt injection if the manifest fields (such as the claim string) were to contain malicious instructions designed to influence the agent after being printed in the script's output.
Ingestion points: scripts/reporting_gate.py reads a JSON file from a path provided as a command-line argument or from standard input.
Boundary markers: The skill requires data to be formatted according to a specific JSON schema, which acts as a structural boundary.
Capability inventory: The processing script is restricted to reading the manifest file and writing a text/JSON report to standard output. It has no capabilities for network operations, subprocess creation, or writing to the filesystem.
Sanitization: The script uses the standard json library for parsing and performs type-checking and presence-validation on required fields before inclusion in the final report.

alterlab-results-transparency