New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Test Ocas Openclaw

by pwrdrvr in pwrdrvr/openclaw-codex-app-server

Regression test the OpenClaw Codex App Server plugin against a live local OpenClaw instance in Telegram or Discord.

MITAuto-check passedAgent Workflows

Install Test Ocas Openclaw

skills CLI
$ npx skills add pwrdrvr/openclaw-codex-app-server --skill test-ocas-openclaw -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install pwrdrvr/openclaw-codex-app-server test-ocas-openclaw --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/pwrdrvr/openclaw-codex-app-server.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/test-ocas-openclaw .claude/skills/test-ocas-openclaw && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
test-ocas-openclaw
GitHub stars
265
Token cost
~4k tokens
SKILL.md length
2,174 words
Files
3 (incl. references)
Skills in repo
3
Repo updated
First seen
Licence
MIT

At a glance

Regression test the OpenClaw Codex App Server plugin against a live local OpenClaw instance in Telegram or Discord.

  • The user wants an end-to-end OCAS manual test pass
  • SKILL.md covers Preconditions, Canonical State Files, Preferred Browser Path and Regression Flow, plus 4 more sections
  • Calls npm
  • Wants to verify binding

What it does

Test Ocas Openclaw is an agent skill from pwrdrvr/openclaw-codex-app-server. Regression test the OpenClaw Codex App Server plugin against a live local OpenClaw instance in Telegram or Discord. Use when the user wants an end-to-end OCAS manual test pass, wants to verify binding, /casresume or /casstatus flags for model/fast/yolo, approvals, plan mode, review, compact, status-card controls, split default-vs-full-access permission behavior, stop-button interrupt handling, rename, or other slash-command behavior, or wants reproducible bug notes from a live chat integration run.

Its SKILL.md is about 4k tokens, which your agent loads only when the skill is triggered. The skill folder holds 4 other files, including reference files (for example `agents/openai.yaml` and `references/agent-browser.md`).

It sits in Agent Workflows, covering Planning, Computer vision and Hooks and plugins. It works with Telegram, Discord and OpenAI. The repository describes itself as: Proof of Concept - Moved to PwrAgent. The licence is MIT.

When your agent uses it

  • The user wants an end-to-end OCAS manual test pass
  • Wants to verify binding
  • /casstatus flags for model/fast/yolo
  • Status-card controls

Example prompts

  • “/test-ocas-openclaw”

What it can do on your machine

Read from SKILL.md and the folder at commit 4dce2e2. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • npm

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Links to these hosts (documentation or services it may open):

    • github.com

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Test Ocas Openclaw loads about 4k tokens when it runs, and up to ~4.3k if it reads all its reference files. Until then it costs about 132 tokens; SKILL.md has 2,174 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~132
When it runs · the whole SKILL.md, loaded when a task matches
~4k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~4.3k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from pwrdrvr/openclaw-codex-app-server at commit 4dce2e2, republished under its MIT licence (© pwrdrvr). 2,174 words, ~3,979 tokens.

Download SKILL.mdSave it as .claude/skills/test-ocas-openclaw/SKILL.md (or your agent's skills folder). This skill also uses 2 other files; get the full folder from GitHub.
name
test-ocas-openclaw
description
Regression test the OpenClaw Codex App Server plugin against a live local OpenClaw instance in Telegram or Discord. Use when the user wants an end-to-end OCAS manual test pass, wants to verify binding, `/cas_resume` or `/cas_status` flags for model/fast/yolo, approvals, plan mode, review, compact, status-card controls, split default-vs-full-access permission behavior, stop-button interrupt handling, rename, or other slash-command behavior, or wants reproducible bug notes from a live chat integration run.

Test OCAS OpenClaw

Use this skill for manual regression passes of this plugin against a real local OpenClaw instance.

Preconditions

  • Confirm the plugin branch under test is linked into the local OpenClaw checkout and OpenClaw is running.
  • Prefer Playwright MCP or Chrome MCP for Telegram Web. It was more reliable than agent-browser in recent passes.
  • Use $agent-browser only as fallback. Read references/agent-browser.md only if you intentionally choose that path.
  • If Telegram Web or Discord is not logged in, stop and ask the user to complete login.
  • Prefer a low-risk Codex thread unless the user asks otherwise. In this repo, discrawl and dupcanon are safe defaults for resume tests.
  • Treat Permissions: toggle as switching between Default and Full Access. If Full Access is unavailable in the current Codex Desktop session, expect the status card to stay in Default mode with an explanatory note.

Canonical State Files

Use these when the UI and the plugin disagree:

  • Core chat binding store: ~/.openclaw/telegram/thread-bindings-default.json
  • Plugin state: ~/.openclaw/openclaw-codex-app-server/state.json

Read them when:

  • a topic looks bound but /cas_status disagrees
  • plan mode appears stuck
  • a questionnaire should exist but Telegram Web does not render it
  • slash replies appear in the wrong chat or topic
  • answer buttons appear but the active pending request does not advance
  • the status card claims a pending permissions migration that never applies after the run finishes

Treat the state files as diagnostic evidence, not as a thing to hand-edit.

Preferred Browser Path

  • Keep the browser anchored to the requested topic or channel for the full pass.
  • For Telegram topics, keep the topic fragment in the URL when possible.
  • If the UI looks stale but local state shows the expected pending request or callbacks, refresh Telegram Web with Cmd+R before calling it a backend bug.
  • Separate browser-navigation failures from OCAS failures. A browser can drift to General or the chat list without proving an OCAS routing bug.

Regression Flow

  1. Establish the target conversation.
  • Stay in the requested group, DM, channel, or topic for the entire pass.
  • If the expected Telegram topic is missing, check whether /cas_resume --sync renamed it. Rename it back or create a fresh test topic if needed.
  • If slash-command routing leaks into General, log that separately and move the browser back to the intended topic before continuing.
  1. Baseline the conversation.
  • Run /cas_status.
  • If the topic is already bound, run /cas_detach first, then confirm the unbound state with /cas_status.
  • For Telegram topics, the reply must appear in the same topic. If it lands in General, record a routing bug.
  1. Resume a thread.
  • Prefer /cas_resume or /cas_resume --all with the normal picker flow.
  • Avoid /cas_resume --all <thread-id> for now. It is a known regression path and may not surface the expected approval UX.
  • Cover at least one flag-driven bind path with /cas_resume --model <name>, /cas_resume --fast or /cas_resume --no-fast, and /cas_resume --yolo or /cas_resume --no-yolo.
  • If the bind path falls back into project or thread pickers, verify the requested model, fast, and yolo settings still apply after the picker choice and final bind approval.
  • When approval appears, use Allow Once unless the user asks otherwise.
  • After binding, send who are you.
  • Expect a Codex response. If the reply sounds like OpenClaw persona text instead of Codex, the bind failed.
  • Run /cas_status again and verify:
    • Binding: active
    • the expected thread id or thread title
    • the expected model
    • Plugin version: 0.0.0 for local dev branch testing when that is the expected version
  1. Verify status-card controls.
  • Use the interactive /cas_status message as the primary control surface.
  • Expect one status message that edits in place rather than a stream of follow-up bot messages.
  • Verify the card includes:
    • Select Model
    • Reasoning: ...
    • Permissions: toggle
    • Compact
    • Stop
  • Verify Fast: toggle only appears when the current model supports fast mode. If the current model does not support fast mode, expect no fast button.
  • Click Select Model.
  • Expect the same message to switch into a model picker.
  • Choose a different model.
  • Expect the same message to return to the status card and show the new model on the Model: line.
  • Run /cas_status again and confirm the new model is still shown.
  • Click the reasoning button.
  • Expect the same message to switch into a reasoning picker.
  • Choose a different reasoning level.
  • Expect the same message to return to the status card and show the new reasoning on the Model: line.
  • Run /cas_status again and confirm the reasoning state persisted independently of the current model.
  • If the current model supports fast mode:
    • click Fast: toggle
    • expect the same message to update Fast mode: immediately
    • run /cas_status again and confirm the fast-mode state persisted
  • If the current model does not support fast mode:
    • use Select Model to switch to a fast-capable model before testing the fast toggle
  • Click Permissions: toggle.
  • If Full Access is available and no turn is active:
    • expect the same message to update Permissions: between Default and Full Access
    • run /cas_status again and confirm the new permissions mode persisted
  • If a turn is active:
    • expect the same message to stay on the current mode and add a note that the requested permissions mode will apply after the current turn ends
    • after the run finishes, run /cas_status again and confirm the pending profile migration actually applied
  • If Full Access is unavailable in the current session:
    • expect the status message to remain on Permissions: Default
    • expect an explanatory note that Full Access is unavailable for the current session
  1. Verify approval rendering in Default mode.
  • Put the conversation back into Permissions: Default before this step.
  • Send this exact prompt as plain text in the bound conversation:
text
I want you to run `npm view dive` and make sure to ask to exit the sandbox as it needs network access.
  • Expect a real execution approval dialog, not a plain text question about whether the user wants approval.
  • Expect approval buttons such as Approve or the platform-equivalent action controls.
  • Expect a code-formatted command area that shows npm view dive.
  • Verify the displayed command is trimmed for presentation and does not leak a shell-launcher wrapper like /bin/zsh -lc ....
  • After verifying the dialog, approve the command with Approve Once or the platform-equivalent approval button so later tests are not blocked by a stale pending approval.
  • If the model only asks a conversational question like Do you want to allow... without rendering execution approval controls, treat that test as invalid and rerun with the exact prompt above.
  1. Verify full-access execution path when enabled.
  • Use the status card to switch into Permissions: Full Access when it is available in the current session.
  • Send this exact prompt as plain text:
text
Run `npm view dive version` and reply with only the version string.
  • Expect the command to run without an approval dialog.
  • Expect a direct result after a short delay rather than an approval request.
  • After the command completes, switch the status card back to Permissions: Default.
  • Run /cas_status again and confirm the card shows Default mode.
  1. Verify long-running review behavior before plan mode.
  • Run /cas_review before /cas_plan.
  • Treat /cas_review as a long-running request. It may take several minutes.
  • Do not interrupt /cas_review by starting /cas_plan, /cas_compact, or other long-running flows while review is active.
  • After starting review, wait 30 seconds at a time and check for output again.
  • Record whether review immediately skips expected desktop-style base and branch questions. That is a behavior gap, but not the same thing as a review failure.
  1. Verify Stop-button behavior on a long-running run.
  • Start a long-running Codex action such as /cas_review, /cas_plan, or a prompt likely to wait on approvals or take more than a few seconds.
  • While the run is active, press Stop on the status card.
  • Expect the active run to interrupt.
  • Expect the same status message to refresh in place after the interrupt.
  • Run /cas_status again and confirm there is no active-run note left behind.
  • If you also want to verify deferred permission migration, start a long-running run, press Permissions: toggle during the run, confirm the pending note, then press Stop and confirm the requested mode applies only after the run ends.
  1. Verify plan mode only after review has finished.
Show full SKILL.md (863 more words)Show less
  • Use /cas_plan <prompt>. Do not test plan mode by sending the raw prompt without the command.
  • Use a short questionnaire prompt first. A breakfast-choice prompt is good because it exercises question rendering and answer submission.
  • Expect:
    • questionnaire text
    • answer buttons
    • question progression when answering
    • final Implement this plan controls
  • If question 1 renders but button clicks or plain-text answers do not advance the active pending request in state.json, record that as a real failure.
  • If plan mode gets stuck, /cas_plan off should exit it.
  1. Verify long-plan truncation.
  • Ask for a final plan longer than 4000 characters.
  • Expect the Telegram preview to truncate and the plugin to attach the full Markdown plan when needed.
  1. Cover other control commands as needed.
  • /cas_resume
    • Verify --model, --fast or --no-fast, and --yolo or --no-yolo paths on direct bind, picker bind, and --new flows.
    • Expect the chosen settings to show up immediately in /cas_status after the bind completes.
  • /cas_mcp
    • Expect a list of configured MCP servers.
  • /cas_skills
    • Expect installed skills.
    • If the response shows both a full text list and buttons for the same skills, record a display bug rather than a blocker.
  • /cas_status
    • Verify in-place message edits for model, reasoning, fast when supported, permissions, compact, and stop controls.
    • Verify the card does not post extra bot messages for each button press unless the platform requires it.
    • Verify /cas_status --model, /cas_status --fast or /cas_status --no-fast, and /cas_status --yolo or /cas_status --no-yolo update the bound conversation settings and refresh the shown status.
  • /cas_compact
    • Expect progress keepalives and a final context-usage report.
  • /cas_rename
    • Verify thread rename and, if requested, --sync topic rename behavior.
  • /cas_diff, /cas_init
    • Verify current forward-or-placeholder behavior instead of assuming full implementation.
    • Treat known-placeholder behavior as neutral, not as a regression, unless the user says the command should already work.
  • Do not look for /cas_model, /cas_fast, or /cas_permissions.
    • Those controls were folded into /cas_resume, /cas_status, and the status card.
  1. Clean up.
  • End by running /cas_detach.
  • Confirm the conversation is no longer bound.
  • Confirm the topic has no lingering pending request in state.json if plan mode, stop, or approvals were exercised.

Result Legend

  • ✅ Pass. Behavior matched what should work now.
  • ❌ Fail. Behavior should work now and did not.
  • ➖ Neutral. The command is known to be unimplemented, placeholder-only, intentionally incomplete, or not configured in this environment.

Use a flat results table while testing:

AreaStatusObservedNotes
/cas_status✅Bound topic reply in-topicInclude thread, model, and plugin version
/cas_resume flags✅ or ❌--model, --fast or --no-fast, --yolo or --no-yolo survive bind flowInclude whether a picker path preserved the settings
Status controls✅ or ❌Same message edits in place for model, reasoning, fast when supported, permissions, compact, and stopRecord extra messages as a bug
/cas_status flags✅ or ❌--model, --fast or --no-fast, --yolo or --no-yolo refresh the binding and status cardNote unsupported fast-model cases separately
Approval dialog✅Real execution approval with trimmed npm view diveApprove after verifying
Full Access execution✅, ❌, or ➖npm view dive version runs without approvalUse ➖ if Full Access is unavailable in the current session
Permission migration✅ or ❌Default vs Full Access persists across /cas_statusNote pending migration if toggled mid-run
Stop button✅ or ❌Active run interrupts and status refreshesRecord whether refresh was in-place
/cas_review✅ or ❌Long-running reviewWait 30s between checks
/cas_plan render✅ or ❌Question text and buttonsSeparate render from answer submission
/cas_plan answer submission✅ or ❌Active request advances or stays stuckCompare against pendingRequests
/cas_diff➖ or ❌Placeholder or real commandUse ➖ if still knowingly unimplemented

Gotchas

  • Telegram Web can go stale. If a simple request appears hung for more than about one minute, and the local state file already shows the pending questionnaire or callbacks, refresh Telegram Web with Cmd+R before calling it a backend bug.
  • Plain text can still route correctly even when slash commands from a topic leak to General. Record both facts separately.
  • A core binding can exist in thread-bindings-default.json while plugin-local state is stale or missing. Compare both files before concluding which side is wrong.
  • Telegram can leave stale questionnaire buttons visible in the DOM. If a click produces No active Codex run is waiting for input. but pendingRequests still shows the active questionnaire unchanged, record that the visible button press did not reach the active request.
  • Starting /cas_plan while /cas_review is still running can interrupt the review. Avoid that sequence during normal regression passes.
  • Permissions migration is edge-triggered on turn completion. If you toggle permissions during an active run, do not expect the profile switch to apply until the run ends or is stopped.
  • A full-access test only proves the split-profile path when the command runs without approval and later Default-mode commands still trigger approval again.

Evidence To Capture

  • Exact command sent
  • Exact chat or topic where the reply appeared
  • Whether the reply was plain text, buttons, or an attachment
  • Whether the status card edited in place or created a new bot message
  • Which permissions mode was selected before the command under test
  • Relevant snippets from state.json and thread-bindings-default.json when UI behavior is suspicious
  • Known-good versus observed behavior
  • The result-table row for each command or flow tested

Formal Bug Notes

If the user wants the findings turned into issues, PR notes, or project-board items, use $project-manager.

© pwrdrvr, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 2 other files (references) in .agents/skills/test-ocas-openclaw of pwrdrvr/openclaw-codex-app-server.

  • SKILL.md
  • agents/openai.yaml
  • references/agent-browser.md

Open the folder on GitHubat commit 4dce2e2

Compare with similar skills

Test Ocas Openclaw next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Test Ocas Openclaw compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Test Ocas Openclaw this skillpwrdrvr/openclaw-codex-app-server265—~4kAutomated safety check: PassMIT
Chat SDKdatabuddy-analytics/Databuddy1.2k—~2.6kAutomated safety check: PassAGPL-3.0
MuseKevPH2026/muse-catch128—~1.6kAutomated safety check: NotesAGPL-3.0
Find AI Consultancyjeremylongshore/tons-of-skills-marketplace2.8k—~4kAutomated safety check: NotesMIT
Codex with ChatGPT Planning LoopXiaoDuoYa/codex-with-chatgpt7.2k—~11kAutomated safety check: NotesMIT
LLM Councilgcpdev/llm-council-skill461—~1kAutomated safety check: NotesMIT

Similar skills

  • Chat SDK

    databuddy-analytics/Databuddy

    Build multi-platform chat bots with Chat SDK (chat npm package).

    1.2k GitHub stars~2.6k tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Muse

    KevPH2026/muse-catch

    Muse · Catch — AI 灵感捕手。浏览器插件 + 任意 Agent（Telegram/WhatsApp/Discord…）一键捕获灵感，AI自动提炼，Web Dashboard浏览。分级LLM路由：Agent内置LLM隐私分析 + TokenRouter云端分布式调用。当用户说「灵感」「捕获」「Muse」「Catch」「记下来」「收藏」「稍后读」时自动触发。

    128 GitHub stars~1.6k tokensUpdated 1 mo ago
    Productivity & AutomationAuto-check: notes

  • Find AI Consultancy

    jeremylongshore/tons-of-skills-marketplace

    A skill your agent uses whenever the user wants to find, shortlist, vet, or enrich US AI/ML/data consulting firms (consultancies) — AI/ML development, MLOps, generative AI / LLM apps (RAG, chatbots…

    2.8k GitHub stars~4k tokensUpdated today
    AI & LLM EngineeringAuto-check: notes

  • Codex with ChatGPT Planning Loop

    XiaoDuoYa/codex-with-chatgpt

    Uses ChatGPT in the browser as the planning and review brain for a Codex session, with Codex keeping all execution and ChatGPT reading the workspace through a bridge.

    7.2k GitHub stars~11k tokensUpdated 9 days ago
    Agent WorkflowsAuto-check: notes

  • LLM Council

    gcpdev/llm-council-skill

    Multi-LLM collaborative brainstorming and planning. An agent skill from gcpdev/llm-council-skill.

    461 GitHub stars~1k tokensUpdated 9 mo ago
    Agent WorkflowsAuto-check: notes

  • Directional Prompting

    kingbootoshi/directional-prompting

    Write prompts, system instructions, agent directives, slash commands, and skill descriptions using two stacked layers — outcome-first (define the destination, success criteria, stopping condition)…

    143 GitHub stars~2.3k tokensUpdated 4 mo ago
    Agent WorkflowsAuto-check passed

More from pwrdrvr/openclaw-codex-app-server

  • Project Manager

    pwrdrvr/openclaw-codex-app-server

    Manage GitHub issues and the GitHub Project board for the current repository, while keeping the local tracker in sync.

    265 GitHub stars~1.5k tokensUpdated 1 mo ago
    Auto-check passed

  • Release

    pwrdrvr/openclaw-codex-app-server

    Plan and publish a GitHub Release in a tag-driven repository.

    265 GitHub stars~2.2k tokensUpdated 1 mo ago
    Auto-check passed

Works with

Categories

Questions about Test Ocas Openclaw

What does Test Ocas Openclaw do?

Regression test the OpenClaw Codex App Server plugin against a live local OpenClaw instance in Telegram or Discord. Test Ocas Openclaw is an agent skill from pwrdrvr/openclaw-codex-app-server. Regression test the OpenClaw Codex App Server plugin against a live local OpenClaw instance in Telegram or Discord.

When should I use Test Ocas Openclaw?

Test Ocas Openclaw fits situations like: the user wants an end-to-end OCAS manual test pass; wants to verify binding; /casstatus flags for model/fast/yolo; status-card controls.

How do I install Test Ocas Openclaw in Claude Code?

Run `npx skills add pwrdrvr/openclaw-codex-app-server --skill test-ocas-openclaw -a claude-code`. Or copy the skill folder (.agents/skills/test-ocas-openclaw in pwrdrvr/openclaw-codex-app-server) into .claude/skills/test-ocas-openclaw in your project. Claude Code loads it when a task matches its description.

How do I install Test Ocas Openclaw in Codex?

Run `npx skills add pwrdrvr/openclaw-codex-app-server --skill test-ocas-openclaw -a codex`. Or copy the skill folder (.agents/skills/test-ocas-openclaw in pwrdrvr/openclaw-codex-app-server) into .agents/skills/test-ocas-openclaw in your project. Codex loads it when a task matches its description.

Can I use Test Ocas Openclaw in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add pwrdrvr/openclaw-codex-app-server --skill test-ocas-openclaw -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/test-ocas-openclaw, .gemini/skills/test-ocas-openclaw, .github/skills/test-ocas-openclaw and .opencode/skills/test-ocas-openclaw in your project.

What does Test Ocas Openclaw need to run?

Going by SKILL.md and its folder, Test Ocas Openclaw needs the command-line tools its instructions call (npm).

Does Test Ocas Openclaw access the network?

SKILL.md names 1 domain. As links in the text: github.com. This is read from the text; nothing was executed.

Is Test Ocas Openclaw safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Test Ocas Openclaw use?

Test Ocas Openclaw is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Test Ocas Openclaw use?

About 4k tokens (SKILL.md is roughly 16k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 351 tokens, read only when the agent opens those files.

What are the alternatives to Test Ocas Openclaw?

Skills that share tags, products or a category with Test Ocas Openclaw: Chat SDK (databuddy-analytics/Databuddy, 1.2k stars), Muse (KevPH2026/muse-catch, 128 stars), Find AI Consultancy (jeremylongshore/tons-of-skills-marketplace, 2.8k stars) and Codex with ChatGPT Planning Loop (XiaoDuoYa/codex-with-chatgpt, 7.2k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Test Ocas Openclaw?

pwrdrvr (a GitHub organization) maintains it in pwrdrvr/openclaw-codex-app-server, which has 265 GitHub stars. The repository holds 3 skills in this directory. The repository was last updated on August 23, 2026.

Source: pwrdrvr/openclaw-codex-app-server on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.