The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill explicitly requests sensitive Atlassian credentials (email and API token) via interactive prompts if they are not found in the execution environment.
[DATA_EXFILTRATION]: The skill retrieves an Atlassian OAuth token and includes it in a curl request to a URL obtained directly from untrusted JIRA ticket JSON. This allows an attacker to exfiltrate the token by providing a malicious URL in a ticket. Furthermore, the skill hardcodes a specific Atlassian tenant URL (tablecheck.atlassian.net) for GraphQL queries, which could lead to credentials being sent to an unintended organization's endpoint.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted data from JIRA tickets, comments, and attachments to define requirements and generate code that is subsequently executed.
Ingestion points: JIRA ticket metadata, descriptions, custom fields, comments, attachment files, and Figma design metadata.
Boundary markers: No protective delimiters or isolation instructions are provided to the sub-agents when processing this external data.
Capability inventory: The skill possesses extensive capabilities including shell command execution (acli, fcli, curl, git), repository file modification, and the ability to run arbitrary test suites.
Sanitization: No validation or sanitization is performed on the incoming ticket data or attachment content before it is processed by the agents.

address-ticket