The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The script scripts/h1_reference.py performs network requests to https://hackerone.com/graphql to retrieve public security report information. HackerOne is a recognized and established service in the security research community.
[COMMAND_EXECUTION]: Documentation within the README suggests the use of various command-line utilities (such as apktool, jadx, and curl) for the purpose of mobile application analysis and endpoint probing. This alignment with the skill's primary focus on security auditing is considered standard behavior for this tool type.
[DATA_EXFILTRATION]: The scripts/secret_scan.py tool is designed to read local files and directories to search for credential patterns (e.g., AWS keys, API tokens). While this involves accessing sensitive local data, it is the core functionality intended for use by security professionals to detect leaks and does not involve sending data to unauthorized third parties.
[PROMPT_INJECTION]: The skill processes untrusted external data fetched from APIs and local file contents, which represents an inherent attack surface for indirect prompt injection.
Ingestion points: Data is ingested from the HackerOne public API and from local files analyzed by the secret scanner.
Boundary markers: The scripts output data as plain text or JSON without specific security delimiters or warnings regarding embedded instructions.
Capability inventory: The skill contains instructions and logic for broad information gathering and system auditing.
Sanitization: Standard data handling and output methods are used; the skill does not implement specific sanitization routines to prevent the execution of instructions that might be embedded within the ingested OSINT data.

offensive-osint