The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Jira issues and comments, which could contain malicious instructions designed to influence the agent's behavior. To mitigate this, the skill establishes strict 'Seven Rules', including Rule 6 which prohibits the agent from changing any wording in the fetched content. It also uses mechanical conversion scripts and automated markdown/JSON validation to ensure data is processed predictably.
Ingestion points: jira-api.mjs (GET requests for issues and comments).
Boundary markers: Conversion via adfToMarkdown and schema validation via check-json.mjs.
Capability inventory: File system writes (writeFileSync), network operations (fetch), and execution of local validation scripts (execFileSync).
Sanitization: Regex-based cleaning and formatting rules implemented in check-markdown.mjs.
[COMMAND_EXECUTION]: The skill uses execFileSync and spawnSync within config.mjs, fetch-issue.mjs, and setup-choices.mjs to execute its internal validation scripts (check-json.mjs, check-markdown.mjs, inspect.mjs). This execution is restricted to the skill's own local environment and tools.
[EXTERNAL_DOWNLOADS]: The skill downloads attachments and wiki pages from Atlassian's official domains. As a well-known service, this is generally considered safe. Additionally, the skill implements an explicit blocklist for executable file types (e.g., .exe, .sh, .bat) to prevent the storage of potentially harmful scripts on the local machine.
[CREDENTIALS_UNSAFE]: The skill demonstrates best practices for credential management by using environment variables (ATLASSIAN_EMAIL, ATLASSIAN_API_TOKEN) and explicitly instructing users never to store tokens in configuration files.

jira-to-local