AgentStack
raul-cgk avatar

raul-cgk

7 listings · 0 installs

Open-source publisher. Listings imported from github.com/raul-cgk — credited to the original author with their license.

↗ github.com/raul-cgk
7 results
Self-run
SKILL

Work Planning

Use when large or uncertain product or engineering work needs a broad, evidence-grounded plan before implementation, especially when shallow assumptions or premature decomposition could distort the approach.

0
3
Free
Self-run
SKILL

Github Triage

Read-only routing of a scoped GitHub issue or pull-request queue. Use when a maintainer needs to know what is ready, needs refinement or owner input, or should be deferred, and when maintainer-orchestrator needs evidence-backed intake.

0
4
Free
Self-run
SKILL

Github Autoreview

Independent review of a GitHub issue, pull request, merged change, or local diff. Use to decide whether behavior is correct, proof is sufficient, and the implementation preserves or improves the codebase's design.

0
7
Free
Self-run
SKILL

Skill Authoring

Create a new Codex or Claude Code skill, or materially redesign an existing skill's reasoning method, boundaries, structure, or downstream contract. Do not use for metadata-only edits or retrospective run evaluation.

0
4
Free
Self-run
SKILL

Maintainer Orchestrator

Codex-native manager control plane for scoped repository work. Use when a root task must coordinate child agents or explicitly authorized durable tasks, track exact candidate state, route independent review, QA, and CI, recover stuck lanes, and surface landing decisions without implementing features itself.

0
5
Free
Self-run
SKILL

Github Issues

Use when the user wants to create one or more GitHub issues from a bounded request, approved plan, or broad parent issue, including decision-complete work that needs decomposition and dependency relationships. Do not use for backlog routing, unresolved planning, implementation, review, or general issue administration.

0
5
Free
Self-run
SKILL

Skill Evaluator

Evidence-backed audit or comparison of completed Codex or Claude Code skill runs. Use to decide what behavior conformed, what failed, why it failed, what should be preserved, and whether supported change candidates should be routed to skill-authoring.

0
7
Free
You've reached the end · 7 loaded