atlassian-readonly-skills
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves and flattens content from Atlassian products (such as issue descriptions, Confluence pages, pull request details, and source code) which are externally controlled and could contain malicious instructions aimed at influencing the agent's behavior.
- Ingestion points: Content is ingested across multiple scripts including
scripts/jira_issues.py,scripts/confluence_pages.py,scripts/bitbucket_files.py, andscripts/jira_search.py. - Boundary markers: While the skill returns data as simplified JSON strings, the individual text fields containing user-generated content are not wrapped in boundary markers or instructions to treat the content as data only.
- Capability inventory: The skill has the capability to perform network requests via the
requestslibrary to configured Atlassian API endpoints and read the local.envconfiguration file. - Sanitization: The implementation extracts specific fields from API responses but does not perform sanitization or filtering on the textual content to detect or neutralize potential prompt injection patterns.
Audit Metadata