codex-prompt-templates

Create, choose, adapt, and validate reusable Codex prompt templates that orchestrate skills, plugins, files, tools, and safety gates. Use when the user asks for massive prompts, prompt templates, reusable Codex prompts, plugin-orchestration prompts, marketplace prompt packs, or turning a prompt into a skill. Triggers: prompt template, reusable prompt, massive prompt, plugin stack prompt, prompt pack, turn this prompt into a skill. Skip when the user only needs a one-off short prompt with no reusable structure.

Route Codex work to the right supercharged workflow. Use when the user asks to improve their Codex setup, choose among skills, combine GitHub repos/MCP servers, reduce tokens, build agent workflows, or decide which local supercharge skill to invoke next. Use the boundary rules here when requests mention agents, evals, costs, guardrails, MCP, daemon/status workflows, or safety.

Detect drift between the Codex Supercharge plugin pack and installed copies under `$HOME/.codex/skills`, including copied `SKILL.md`, `references/`, and `scripts/` payloads. Use when behavior differs across machines/sessions or after editing plugin skills, references, or helper scripts without reinstalling.

Verify current GitHub and agent-skill ecosystem evidence before using trend, popularity, star, topic, or public-discoverability claims in Codex skill and plugin work. Use when an automation asks for trend evidence, GitHub discovery, repo stars/forks/activity, launch topics, marketplace positioning, or latest agent-skill/MCP ecosystem claims. Skip when the task only needs local source synthesis or a single trusted repo brief.

Add lightweight local observability to Codex/agent workflows. Use when tracking local runs, traces, evals, cost, token usage, cache efficiency, model comparisons, failure rates, or claudetop-style monitoring. Skip when designing production gateway policy, staged guardrails, shadow traffic, model budgets, or trace-to-regression loops; use `$auto-skill-build-agent-reliability-loop`.

Run closed-loop evaluators for Codex Supercharge artifacts and write feedback state for future Codex sessions, subagents, and automations. Use after creating, editing, reviewing, grading, or validating Codex Supercharge skills, references, scripts, metadata, or public docs. Triggers: grade Supercharge edits, closed-loop feedback, evaluator handoff, subagent status, automation feedback. Skip when the task is a one-off repo edit outside Codex Supercharge.

Run a deterministic Codex Supercharge maintenance pass (skill audit, plugin discovery audit, helper smoke checks, quick validation, optional install sync plus installed-payload drift verification). Triggers: maintenance pass, validate skills, run skill audit, validate helper scripts, resync installed skills. Skip when you are not changing Codex Supercharge skills/scripts.

Create or use lightweight persistent project memory for Codex. Use when the user wants Codex to remember decisions, naming, architecture rules, progress, experiment results, or prior debugging across sessions. Also use with memsearch-style markdown memory.

Design production agent reliability loops across traces, evals, guardrails, gateways, model budgets, and shadow tests. Use when shipping AI/agent workflows that need simulate/evaluate/protect/monitor/optimize structure, gateway policy, eval suites, guardrail rollout, or trace-to-regression feedback. Skip when the task only needs a local Codex run ledger; use `$agent-observability` instead.

Append and summarize a lightweight local `.agent-runs/runs.jsonl` ledger (claudetop-inspired) so Codex sessions build durable observability over time. Use at the end of a significant task, after running tests, or when you want a quick “what did we do lately?” dashboard. Skip when the user only wants a one-off fix and does not care about tracking runs.

Design a repeatable MCP conformance, health-check, and tool-surface regression harness for local or CI use. Use when the user asks for MCP CI, protocol or OAuth conformance, server doctor checks, tool-surface diffing, JUnit/XML reports, or release gates for an MCP server. Requires an MCP endpoint or stdio launch command; skip when the task is only ad-hoc manual inspection where `$mcp-server-lab` is enough.

Use memsearch as an on-demand semantic index for project markdown memory (search → expand → optional deep drill). Use when the user asks to recall past decisions, prior debugging, or “have we seen this before?” and a `.memsearch/memory/` store exists (or the user wants to set one up). Skip when the question is only about current code state (use Read/Grep) or when the user explicitly does not want semantic indexing / installs.

Synthesize multiple local `sources/` clones into workflow families before creating skills, references, or scripts. Use when doing source-pattern synthesis, comparing several GitHub-derived repos, deciding which cloned patterns deserve durable Codex artifacts, or avoiding duplicate generated skills. Requires local source clones plus `source_pattern_matrix.py`. Skip when you only need a one-repo brief or a metadata-only source inventory.

Generate and refresh a metadata-only inventory of local `sources/` shallow clones (size, git remote, HEAD, commit date, license filename) and write it into `plugins/codex-supercharge/references/source-repos.md`. Use when you want a fast “what sources do we have?” view, to control disk growth, or before doing a source-pattern synthesis pass. Skip when there is no `sources/` directory or you are not working with local clones.

Safely inspect and adapt third-party agent/Codex/Claude skill packs before installation. Use when importing external skills, prompt packs, slash commands, Skill Forge candidates, OpenDesign-style catalog stubs, or repos with SKILL.md files, commands, hooks, or installers. Triggers: import skills, install skill pack, evaluate third-party skill, prompt-pack safety, adapt Claude skill to Codex. Skip when authoring a new first-party skill from scratch; use `$skill-pack-factory` and `$automation-skill-forge`.

Prevent secret leakage when adding or reviewing MCP server configs. Use when you see `mcpServers` JSON, `.env`, config snippets, tool docs, or logs that might contain API keys/tokens. Produces a redacted config + a short secret hygiene report.

Turn automation runs into one or more specific Codex skill candidates. Use when an automation should produce reusable skills, stronger prompts, trend workflows, GitHub-ready skill artifacts, or generated-skill naming rules. Triggers: automation to skill, make this into a skill, generated skill pack, many skills from GitHub repos, trend-backed skill, skill naming regex. Skip when the task only needs a one-off fix or a broad brainstorm.

Use graph-backed or semantic codebase intelligence before editing. Use when tasks need symbol search, blast radius, affected tests, architecture health, dependency paths, large-codebase context, Roam, CodeAlive, GraphRAG, or MCP code search. Skip for tiny scripts where rg/read is enough.

Run a controlled THINK-EXECUTE-REFLECT loop for coding, research, or ML experiments. Use when the user wants iterative improvement, benchmark-driven changes, autonomous experiment planning, low-cost monitoring, or progress reports without uncontrolled agent drift. Skip production release loops that require gateway policy, trace-to-regression evals, guardrail rollout, budgets, or shadow/canary traffic; use `$auto-skill-build-agent-reliability-loop`.

Build compact Codex task context and avoid token blowups. Use when work spans large repos, repeated sessions, long logs, many files, expensive models, or the user mentions limits, context, tokens, compression, summaries, or keeping Codex fast.

Review MCP servers, agent tools, install scripts, or fresh GitHub repos for security and trust risk. Use when tools request shell/filesystem/network access, credentials, browser/session control, messaging accounts, package installs, or destructive write permissions. Skip pure CI/conformance harness design after trust is established; use `$auto-skill-build-mcp-conformance-harness`.

Inspect, test, debug, or integrate MCP servers for Codex workflows. Use for MCPJam, MCP Inspector, stdio/HTTP MCP configs, tools/resources/prompts, Apps SDK widgets, OAuth checks, conformance, evals, or creating a local MCP integration plan. Skip when the task is specifically to design repeatable CI conformance gates or release diffs; use `$auto-skill-build-mcp-conformance-harness` instead.

Decide when to use cheaper, local, or frontier models and how to measure cost tradeoffs. Use for token limits, model routing, cost-conscious Codex work, DeepSeek/Claude/OpenAI comparisons, prompt caching, or agent workflow economics. Skip when cost routing is enforced by a production gateway, budget policy, fallback policy, or live alerting; use `$auto-skill-build-agent-reliability-loop`. Avoid unverifiable savings claims.

Use the local OpenDesign design-system library when building, redesigning, or reviewing UI. Triggers: making UI, frontend design, SaaS UI, dashboard, landing page, mobile app, design system, make it look better. Skip when the task is backend-only or the user supplied a stricter design spec.

Convert OpenAPI/Swagger APIs into Codex-friendly MCP/tool workflows and generated skills. Use when the user wants to expose an API to agents, create a skill from API docs, use toolcast-style conversion, or design safe API tool schemas. Triggers: OpenAPI to MCP, Swagger to tools, API tool plan, generate MCP server from API spec. Skip when there is no stable spec or the user only needs one ad-hoc HTTP request.

Inspect a GitHub or local repository and produce a compact Codex-ready architecture/task brief. Use for repo onboarding, weekend project selection, fork analysis, setup discovery, extension ideas, or turning a source repo into a useful build plan.

Build a local markdown wiki for a repo, research folder, or cloned source collection. Use when the user wants living docs, repo explainers, source synthesis, citations, cross-links, or llmwiki-style knowledge accumulation.

Safely clone, inspect, and triage fresh GitHub repos without overusing disk or running risky code. Use when the user asks to clone many repos, evaluate bleeding-edge tools, install MCP servers, run unknown scripts, or harvest patterns from public repos.

Create non-overlapping Codex skills or a local plugin from repos, workflows, PDFs, APIs, or repeated tasks. Use when the user asks to make many skills, convert GitHub learnings into skills, build a skill catalog, or package a workflow into reusable Codex instructions.

Audit Codex skills and plugins for trigger clarity, duplication, context bloat, missing references, unsafe instructions, and validation gaps. Use before installing, publishing, or relying on generated skills.

Set up a practical Codex workflow for a new day, project, or repo. Use when the user wants an amazing workflow setup, a repeatable launch checklist, aligned skills, project memory, low-token defaults, daemon/process control, multi-agent coordination, or a personal Codex operating rhythm.