prag-man
UserPrivacy-first agent skills and local CLI for shaping, routing, researching, prototyping, and improving work across Codex, Claude Code, and Cursor.
Categories
Indexed Skills (8)
pragman-analyze
Use after a project, decision, incident, experiment, release, or workflow to explain outcomes from evidence and choose the highest-leverage next improvements.
pragman-init
Use for first-time Pragman setup, major reconfiguration, broken or shadowed installations, or deciding how Pragman should improve an existing agent workflow; skip it for routine routing after setup is healthy.
pragman-prototype
Use when a product, workflow, or UI decision needs a disposable clickable artifact to test its riskiest assumption before production implementation.
pragman-research
Use when a decision or deliverable needs current external, technical, market, competitive, internal-workspace, or multi-source evidence tied to an actionable recommendation.
pragman-router
Use as the adaptive front door for non-trivial, ambiguous, multi-domain, operational, or explicitly routed work to choose the smallest safe capability sequence; skip it for obvious bounded requests.
pragman-shape
Use when an idea or request is vague, oversized, premature, or low-evidence and must become a bounded decision, experiment, prototype, issue, spec, or implementation-ready task.
pragman-unfck
Use when agentic work feels slow, chaotic, repetitive, over-scoped, unreliable, expensive, stuck across sessions, or mismatched to the user's workflow, and the fix must be evidence-backed and reversible.
pragman-workspace
Use when creating, inspecting, linking, changing, or validating reusable personal, company, client, venture, or project context; keep one-off facts in the current request.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.