The Agent Skills Directory

[CREDENTIALS_SAFE]: The skill manages Jira authentication via a dedicated setup script (jira-setup.py). It stores credentials in ~/.env.jira or ~/.jira/profiles.json using restricted filesystem permissions (0600), ensuring that secrets are only readable by the owner. Additionally, a utility in scripts/lib/errors.py automatically redacts sensitive patterns such as bearer , token=, and Authorization: headers from error messages before they are displayed to the user.
[EXTERNAL_DOWNLOADS]: The jira-attachment.py script provides secure file download capabilities. It strictly validates that output paths are constrained within the current working directory to prevent path traversal attacks and ensures that download targets utilize HTTPS.
[DATA_EXFILTRATION]: The skill includes defensive checks to prevent the accidental leakage of Jira credentials to external domains. In scripts/core/jira-attachment.py, the resolve_attachment_url function verifies that attachment download URLs match the configured Jira host's origin (host and port) before initiating an authenticated request.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to ingest data from Jira issues (descriptions and comments), the skill is exposed to indirect prompt injection. It mitigates this by using structural formatting and clear boundary markers in scripts/lib/render.py (e.g., Description: headers and --- [created] author --- separators) to help the AI agent distinguish between instructions and ingested data.
[COMMAND_EXECUTION]: The skill utilizes local Python scripts executed through uv run. Static analysis identified a subprocess.run() call in scripts/core/jira-validate.py, which was manually verified to be a benign operation used to check the version of the uv package manager during environment validation.

jira-communication