The Agent Skills Directory

Secure Credential Handling: The skill manages Jira API tokens and Personal Access Tokens (PAT) via environment variables or a protected local file (~/.jira.env). It implements a specific protocol to ensure the agent does not ask for or display these credentials in the chat interface, and checks file permissions on POSIX systems to ensure owner-only access.
Secret Redaction: The implementation includes a comprehensive redaction mechanism in scripts/jira.py that identifies and masks sensitive patterns such as 'Authorization' headers, API tokens, and session cookies in logs, error messages, and standard output.
Hardened Network Operations: The skill uses a custom HTTP opener that enforces TLS for non-loopback connections and explicitly disables automatic redirects. This prevents potential credential theft via malicious redirects to untrusted hosts.
Input Sanitization and Size Limits: To prevent resource exhaustion or malformed requests, the skill enforces size limits on input payloads and response bodies, and validates identifiers like issue keys and project keys using regular expressions.
Indirect Prompt Injection Surface: The skill ingests data from external Jira instances (issue summaries, comments). While this presents a potential surface for indirect prompt injection, the skill's documentation explicitly warns that this content should be treated as untrusted data.
Ingestion points: Jira REST API responses (issue fields, comments) processed in scripts/jira.py.
Boundary markers: The skill primarily outputs structured JSON, providing natural boundaries for data.
Capability inventory: The skill can perform network requests to Jira and write issue updates, transitions, and comments.
Sanitization: The skill performs extensive redaction of secrets, though it does not filter natural language instructions from retrieved content.

jira