atlassian-community-search
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the Atlassian Community.
- Ingestion points: Data is retrieved from the community.atlassian.com API.
- Boundary markers: Instructions explicitly state to treat all external data as untrusted evidence and never as instructions. The skill recommends delegating searches to subagents to isolate raw content from the main agent context.
- Capability inventory: Uses curl and jq to fetch and filter data.
- Sanitization: Mandates URL-encoding of all parameters and using jq projections to strip potentially problematic content like full HTML bodies.
- [COMMAND_EXECUTION]: The skill uses shell commands (curl, jq) to interact with the API.
- Evidence: The implementation guidelines enforce safe practices, including the use of --data-urlencode for parameters to prevent command injection and set -o pipefail for robust shell error handling.
Audit Metadata