skills/annals-ai/ah-cli/ah-a2a/Gen Agent Trust Hub

ah-a2a

Pass

Audited by Gen Agent Trust Hub on May 18, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill enables the transfer of local files to remote agents via the '--upload-file' flag. It also includes a '--base-url' flag in the 'ah chat' command, which permits the redirection of communication to arbitrary external servers.
  • [COMMAND_EXECUTION]: The skill's core functionality relies on executing the 'ah' CLI utility to interact with the A2A network.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it is designed to ingest and process outputs from remote agents.
  • Ingestion points: Responses from remote agents are received through 'ah call' and 'ah chat' commands (SKILL.md).
  • Boundary markers: No explicit delimiters are used to separate remote agent content from the local context.
  • Capability inventory: The agent can execute shell commands, upload files to remote targets, and write files to the local system.
  • Sanitization: There is no documentation of sanitization or filtering for data retrieved from the network.
Audit Metadata
Risk Level
SAFE
Analyzed
May 18, 2026, 04:07 AM