ask-codex
Ask Codex from a non-Codex harness through a full ACP session for an independent answer or explicitly scoped implementation.
Skill by jesse-merhi
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/jesse-merhi-ask-codex
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/ask-codex
What's inside
- Skill name
ask-codex- Description
- Ask Codex from a non-Codex harness through a full ACP session for an independent answer or explicitly scoped implementation.
- License
- MIT
- Instructions
- 51 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Ask Codex Use a full Codex session through ACP. This is not a subagent and must not be implemented with the current harness's in-chat delegation tools. Invoke this skill only when the current user explicitly asks to run `$ask-codex` or explicitly says to ask Codex. Never auto-select it for review, planning, a second opinion, delegation, or implementation. Mentioning, discussing, or linking this skill is not authorization to run it. ## Workflow 1. Resolve `<skill-dir>` to this skill's directory. 2. Give Codex a self-contained brief: objective, checkout, relevant files, constraints, expected output, whether writes are authorized, and the requirement to archive its own external task before finishing. 3. For advice, review, or planning, run the read-only wrapper: ```sh <skill-dir>/scripts/ask-codex read "<self-contained prompt>" ``` 4. Use write mode only when the user authorized implementation and the brief names the owned scope: ```sh <skill-dir>/scripts/ask-codex write "<self-contained prompt>" ``` 5. Record the created Codex task or session ID, treat the result as another full session's work, inspect its evidence, and validate any edits in the originating session before reporting completion. 6. A one-shot session is temporary. Require the spawned Codex session to archive itself with `set_thread_archived` before it finishes. After its result has been captured, verify that it is archived. If self-cleanup did not happen, archive the exact task in guaranteed caller cleanup on success, failure, cancellation, timeout, or early stop. Use `set_thread_archived` when the Codex app tool is available; otherwise run `codex archive <id>`. If archival fails, report the exact ID instead of silently leaving the task behind. The wrapper uses the current Codex configuration; do not hard-code a model. Use a fresh one-shot ACP session by default. Continue a persistent named session only when the user explicitly asks for an ongoing cross-harness conversation. Keep that session while the conversation is active, then archive it when the user ends the conversation. If ACP or Codex authentication fails, report the exact failure. Do not silently replace the requested Codex session with a subagent or self-answer.
From jesse-merhi/skills, MIT.
More from jesse-merhi/skills
- 0.569score
just-do-it
jesse-merhi
Take one well-defined, cohesive change from its current implementation or delivery checkpoint through exact-head code review, proof, CI, and a reviewer-ready…
SkillBFreshness: B★ 4Claude Code +9 - 0.566score
ask-claude
jesse-merhi
Ask Claude from a non-Claude harness through a full ACP session for an independent answer or explicitly scoped implementation.
SkillBFreshness: B★ 4Claude Code +9 - 0.566score
feedback-hardening
jesse-merhi
Evidence-backed user corrections or self-detected mistakes revealing reusable agent failures: delegate one systemic-fix recommendation, get approval, then…
SkillBFreshness: B★ 4Claude Code +9 - 0.565score
cleanup
jesse-merhi
Discover and remove local development artifacts that belong to finished or abandoned work, including worktrees, branches, stashes, processes, watchers…
SkillBFreshness: B★ 4Claude Code +9 - 0.563score
grilling
jesse-merhi
Stress-test a plan, decision, or idea through relentless questions and explicit grill requests.
SkillBFreshness: B★ 4Claude Code +9 - 0.563score
atlassian-cloudid-jira
jesse-merhi
Query Jira, JPD, and Confluence through local Rovo Dev MCP, including site IDs, JQL, CQL, and boards.
SkillBFreshness: B★ 4Claude Code +9 - 0.563score
finding-discipline
jesse-merhi
Confirm actionable review findings, deduplicate root causes, and exclude nits, vague risks, and style notes.
SkillBFreshness: B★ 4Claude Code +9 - 0.562score
model-writing-guides
jesse-merhi
Select model-specific writing guidance for another skill, including exact coverage checks, same-family fallback, and one task-wide stale-coverage notice.
SkillBFreshness: B★ 4Claude Code +9 - 0.562score
design
jesse-merhi
Route UI design work among production visual design, interaction and motion implementation, read-only motion review, and explicit multi-variant prototyping.
SkillBFreshness: B★ 4Claude Code +9 - 0.562score
design-technical-diagrams
jesse-merhi
Create or refine architecture, lifecycle, sequence, trust-boundary, decision, threat-model, SVG, or HTML diagrams that explain system relationships.
SkillBFreshness: B★ 4Claude Code +9 - 0.562score
handoff
jesse-merhi
Transfer the current work to a fresh full agent session, placing related work beside the current session and unrelated asides separately.
SkillBFreshness: B★ 4Claude Code +9 - 0.559score
improve-codebase-architecture
jesse-merhi
Review architecture for module depth, interfaces, locality, dependencies, and testability.
SkillBFreshness: B★ 4Claude Code +9
Signals
- Stars
- 4
- Forks
- 2
- Open issues
- 0
- Last commit (this folder)
- Sep 4, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/jesse-merhi/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from jesse-merhi
- 0.569score
just-do-it
jesse-merhi
Take one well-defined, cohesive change from its current implementation or delivery checkpoint through exact-head code review, proof, CI, and a reviewer-ready…
SkillBFreshness: B★ 4Claude Code +9 - 0.566score
feedback-hardening
jesse-merhi
Evidence-backed user corrections or self-detected mistakes revealing reusable agent failures: delegate one systemic-fix recommendation, get approval, then…
SkillBFreshness: B★ 4Claude Code +9 - 0.566score
ask-claude
jesse-merhi
Ask Claude from a non-Claude harness through a full ACP session for an independent answer or explicitly scoped implementation.
SkillBFreshness: B★ 4Claude Code +9 - 0.565score
cleanup
jesse-merhi
Discover and remove local development artifacts that belong to finished or abandoned work, including worktrees, branches, stashes, processes, watchers…
SkillBFreshness: B★ 4Claude Code +9 - 0.563score
grilling
jesse-merhi
Stress-test a plan, decision, or idea through relentless questions and explicit grill requests.
SkillBFreshness: B★ 4Claude Code +9 - 0.563score
atlassian-cloudid-jira
jesse-merhi
Query Jira, JPD, and Confluence through local Rovo Dev MCP, including site IDs, JQL, CQL, and boards.
SkillBFreshness: B★ 4Claude Code +9
Cite this page
ask-codex skill. Agent Mart. https://agentmart.dev/skills/jesse-merhi-ask-codex (last updated 2026-10-24)