The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/github_daily_report.py executes the gh (GitHub) CLI using subprocess.run. This is an intended function of the skill to retrieve repository data (commits, pull requests, issues) for the authenticated user. The arguments are constructed using internal logic with specific JSON queries, which prevents arbitrary command injection.
[EXTERNAL_DOWNLOADS]: The skill performs several legitimate network requests to populate its data:
Fetches real-time exchange rates from cdn.moneyconvert.net in scripts/get_rate.py.
Retrieves New Zealand public holiday information from the Nager.Date API at date.nager.at in scripts/get_holidays.py.
Aggregates news headlines and summaries from reputable RSS feeds (e.g., TechCrunch, BBC, MIT Technology Review) and well-known aggregators like rsshub.app in scripts/fetch_news.py.
[PROMPT_INJECTION]: An indirect prompt injection surface is present because the skill ingests and processes untrusted external data.
Ingestion points: scripts/fetch_news.py (external RSS feeds) and scripts/github_daily_report.py (GitHub commit messages and issue titles).
Boundary markers: None. The skill assembles the raw text from these sources for the agent to translate and summarize without explicit delimiters to isolate untrusted content.
Capability inventory: The skill can execute the gh CLI via subprocess.run to read data from the user's GitHub account.
Sanitization: scripts/fetch_news.py performs basic HTML tag removal using regex before processing the text.
Assessment: The risk is minimal as the ingested data (news and commit logs) is formatted for human reading and the agent is instructed to perform a specific translation and summarization task.

devpilot-daily-toolkit