jira-to-speckit
Pass
Audited by Gen Agent Trust Hub on Nov 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data from Jira issues, which is a known attack surface.
- Ingestion points: Fetches issue summaries, descriptions, and comments from a remote Jira API.
- Boundary markers: The
SKILL.mdfile contains explicit instructions to treat all fetched content as data, not instructions, and to move suspected injection attempts to an 'Open questions' section instead of following them. - Capability inventory: The agent has access to
bash,view, andcreate(file writing) tools. - Sanitization: The skill implements a compaction pipeline that normalizes ADF/wiki markup to plain text and enforces character budgets, which can help strip complex hidden payloads from the prompt context.
- [COMMAND_EXECUTION]: The skill uses the
bashtool to perform network requests. - Evidence: Execution of
curlcommands to interact with the Jira REST API. The skill specifically mandates a secure invocation pattern (printf ... | curl -K -) that prevents the Jira API token from being visible in system process listings (e.g., viaps), protecting against credential harvesting by other processes on the same host. - [EXTERNAL_DOWNLOADS]: The skill performs legitimate network operations to fetch ticket information.
- Evidence: Connection to the host defined in the
JIRA_URLenvironment variable. A guardrail is included to ensure credentials are never sent to external browse URLs provided as input, only to the configured Jira instance. - [CREDENTIALS_UNSAFE]: The skill manages authentication using API tokens.
- Evidence: While tokens are required, the skill explicitly forbids hardcoding credentials or requesting them from the user in chat. It mandates using a project-local
.envfile and uses a safe parsing loop (while readinstead ofsource) to load them without executing potential malicious shell code in the environment file.
Audit Metadata