The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes multiple shell commands to perform deep repository analysis and maintain its memory state. Specifically, it uses git rev-parse, git log, git diff, git blame, and rg (ripgrep) to inventory code capabilities, identify code churn, and track changes since the last ideation run. These commands are executed locally to provide the agent with a technical baseline of the project.
[DYNAMIC_EXECUTION]: During the final reporting phase (Phase 9), the skill generates and executes a bash script to handle directory creation (mkdir -p) and to dynamically resolve a unique filename for the report. The script iterates through potential filenames based on the current date and the presence of existing files to avoid overwriting previous reports.
[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection by fetching and processing data from external, user-generated platforms. In Phase 4 (Market Scan), a subagent is tasked with searching the web and fetching content from sources like Reddit, Hacker News, and GitHub issues to identify market gaps and user sentiment. This untrusted external data is subsequently used in the generation and convergence phases. The skill implements a mitigation strategy in Phase 7 (Critic), where a fresh agent reviews the shortlist and insights for validity against the codebase and primary sources.

ideate