The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell and Node.js scripts to coordinate its workflow. It leverages system utilities such as gcalcli to fetch calendar data and the GitHub CLI (gh) to retrieve developer profile information.
[EXTERNAL_DOWNLOADS]: Conducts automated web research on meeting attendees by utilizing the web_search tool and GitHub CLI to extract professional data from public platforms like LinkedIn and GitHub. This is a core function of the skill.
[DATA_EXFILTRATION]: Sends generated research briefs to the user's verified email address via the gmail skill and optionally to a user-configured Slack webhook using curl. These network operations target well-known services and are restricted to delivering content to the user.
[PROMPT_INJECTION]: Processes external content, including calendar descriptions and web search results, by interpolating them into AI prompts for summary generation. This identifies a surface for indirect prompt injection, where malicious text in a public profile could attempt to influence the brief's output, though risks are mitigated by the localized nature of the delivery.
Ingestion points: Meeting titles and descriptions from scripts/check_calendar.sh and web/GitHub data gathered in scripts/research_person.js.
Boundary markers: Prompt templates in scripts/generate_brief.js and scripts/research_person.js use markdown headings and JSON stringification to delineate external data from instructional text.
Capability inventory: Execution of local bash and node scripts, local file system writes for caching research and logging sent briefs, and access to the gmail and web_search capabilities.
Sanitization: The skill includes scripts/format_for_slack.sh to sanitize and reformat markdown for Slack compatibility, while the prompt architecture uses structured templates for data handling.

meeting-brief