The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes Jira issue descriptions to extract milestone dates, which are then used to update project fix versions. This ingestion of untrusted external data creates a surface for indirect prompt injection.
Ingestion points: The function lookup_release_doc in scripts/_release_dates.py fetches and processes the description field from Jira issues.
Boundary markers: The skill does not use explicit delimiters or instructions to treat the ingested data as untrusted content.
Capability inventory: The skill has the capability to write to the Jira API, specifically creating, updating, and deleting versions, which could be misused if malicious instructions are embedded in issue descriptions.
Sanitization: Data is parsed using a specific milestone extraction logic, but the raw text is handled within the agent's workflow.
[DYNAMIC_EXECUTION]: The script scripts/_release_dates.py dynamically modifies the Python module search path at runtime to load dependencies.
Evidence: It uses sys.path.insert(0, str(_jira_api_scripts)) where the path is computed relative to the script's location. This is used to import the adf_milestones module from a sibling skill directory. While standard for this repository's structure, dynamic path modification is a category of runtime code loading.
[DATA_EXFILTRATION]: The skill accesses sensitive local files to retrieve authentication credentials for Jira.
Evidence: scripts/_auth.py searches for and reads credentials from .jira-token and ~/.config/.jira/.config.yml. This access is a necessary part of the tool's primary purpose to authenticate the user against the Jira REST API.
[COMMAND_EXECUTION]: The documentation in SKILL.md and workflows/sync-fixversions.md instructs the user and agent to execute multiple shell commands using uv run to perform synchronization and planning tasks.

rhdh-release-fixversions