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

Provider Chat Smoke

by Porabuild in Porabuild/Poracode

Smoke test real Poracode provider chat threads and ACP sessions end to end.

Apache-2.0Auto-check passedTesting & QA

Install Provider Chat Smoke

skills CLI
$ npx skills add Porabuild/Poracode --skill provider-chat-smoke -a claude-code

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

GitHub CLI
$ gh skill install Porabuild/Poracode provider-chat-smoke --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/Porabuild/Poracode.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/provider-chat-smoke .claude/skills/provider-chat-smoke && 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
provider-chat-smoke
GitHub stars
115
Token cost
~4.2k tokens
SKILL.md length
1,422 words
Files
2
Skills in repo
21
Repo updated
First seen
Licence
Apache-2.0

At a glance

Smoke test real Poracode provider chat threads and ACP sessions end to end.

  • Works in 4 steps: Inspect the diff, generate the… → Separate gates into live (an… → Create a fresh thread per provider. Use… → …
  • Validating Qwen Code
  • SKILL.md covers Plan coverage before launch, Required live chat matrix, ACP connection analysis and Deterministic ACP branch matrix, plus 1 more section
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Provider Chat Smoke is an agent skill from Porabuild/Poracode. Smoke test real Poracode provider chat threads and ACP sessions end to end. Use when validating Qwen Code, Kimi Code, or another structured/ACP provider; testing chat turn handling, steer, Stop, question or permission tools, live model changes, session resume, ACP handshake/capabilities, or provider-chat regressions.

Its SKILL.md is about 4.2k tokens, which your agent loads only when the skill is triggered. The skill folder holds 2 other files (for example `agents/openai.yaml`).

It sits in Testing & QA, covering QA and bug reports. It works with Kimi and Qwen. The repository describes itself as: One window for all your AI coding agents. Run Claude, Codex, OpenCode, Gemini, Antigravity, Cursor, and Copilot side-by-side. Terminal and chat, any layout. The licence is Apache-2.0.

When your agent uses it

  • Validating Qwen Code
  • Another structured/ACP provider
  • Testing chat turn handling
  • Permission tools

Example prompts

  • “/provider-chat-smoke”

Workflow steps

4 steps, taken from the first numbered list in SKILL.md.

  1. Inspect the diff, generate the integration smoke plan, and record every provider/presentation surface in scope.
  2. Separate gates into live (an authenticated provider and its real ACP server) and deterministic (mock/unit coverage for protocol branches…
  3. Create a fresh thread per provider. Use short marker prompts such as QWEN_SMOKE_OK and KIMI_SMOKE_OK; never use repository-changing prompts.
  4. Capture screenshots, runtime items, final thread state, and the first three console/runtime errors. Do not call the suite successful while…

What it can do on your machine

Read from SKILL.md and the folder at commit e56fcc8. 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

    No scripts in the folder and no shell commands in SKILL.md.

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

  • Network

    No URLs in SKILL.md.

    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

Provider Chat Smoke loads about 4.2k tokens when it runs. Until then it costs about 85 tokens; SKILL.md has 1,422 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~85
When it runs · the whole SKILL.md, loaded when a task matches
~4.2k

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 Porabuild/Poracode at commit e56fcc8, republished under its Apache-2.0 licence (© Porabuild). 1,422 words, ~4,168 tokens.

Download SKILL.mdSave it as .claude/skills/provider-chat-smoke/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
provider-chat-smoke
description
Smoke test real Poracode provider chat threads and ACP sessions end to end. Use when validating Qwen Code, Kimi Code, or another structured/ACP provider; testing chat turn handling, steer, Stop, question or permission tools, live model changes, session resume, ACP handshake/capabilities, or provider-chat regressions.

Provider Chat Smoke

Use this skill with interactive-testing. Run the ordinary isolated-app workflow from that skill first; use real provider credentials only for the safe live gates below. Keep the scope to a disposable project, never approve a write, and preserve the smoke profile for inspection.

Plan coverage before launch

  1. Inspect the diff, generate the integration smoke plan, and record every provider/presentation surface in scope.
  2. Separate gates into live (an authenticated provider and its real ACP server) and deterministic (mock/unit coverage for protocol branches that cannot safely be driven against an external provider).
  3. Create a fresh thread per provider. Use short marker prompts such as QWEN_SMOKE_OK and KIMI_SMOKE_OK; never use repository-changing prompts.
  4. Capture screenshots, runtime items, final thread state, and the first three console/runtime errors. Do not call the suite successful while a required gate is unresolved.

Required live chat matrix

Run each applicable row for every requested provider. Retry a timing-sensitive row with a longer harmless response if the provider finishes before the control can be used.

GateDrive through the real UIRequired evidence
Launch and first turnChoose the provider/model, submit a marker prompt, and wait for idle.User row, matching assistant marker, no error state.
Image attachment turnPaste/attach an image in the composer of the first turn (marker prompt referencing it), wait for idle.Image chip on the user row, the assistant actually reads/acknowledges the image content (providers with readsImageAttachmentsFromHost: false read the localized file via their own tools), and no media-rejection error state.
Commands, skills, pluginsType / in the composer: provider-native slash commands, skills, and plugin contributions must appear.Provider-declared commands/skills render (from the slashCommands capability or a provider-owned catalog via reportsSkillCatalog); picking one submits its invocation. If the provider exposes none, report that with the capability snapshot — do not accept a bare generic list as PASS.
Sub-agent renderingTrigger a sub-agent run (a prompt that spawns subtasks, e.g. parallel research), wait for completion.subagent_*/Task-style tool calls render as sub-agent rows in the composer and the thread panel (never as plain tool rows), and child activity stays inside the agent row/overlay instead of flooding the main timeline.
Follow-upSend a second marker in the same thread.Ordered second user/assistant pair.
Structured questionAsk the agent to invoke its question tool with two choices; submit a benign answer.Question dock, selected answer, turn resumes and completes. Submit must be disabled until every question is answered — an empty submit must never reach the supervisor.
PermissionAsk for exactly pwd (or another read-only command), wait for the approval dock, then choose Reject/Deny.Approval details, declined tool/command item, agent confirms it did not run (for filesystem-verifiable actions, check the file was not created).
SteerStart a long harmless response; while status is working, send a replacement instruction with a marker.Working state, pending-steer strip or equivalent, original turn cancels, replacement marker completes.
StopStart a long harmless response, press Stop response promptly, and wait for terminal state.Stop control accepted, no endlessly-working thread, cancellation/idle/error outcome recorded.
Mid-thread model changeBetween completed turns, change to a different model offered by the same provider, then send a marker prompt.Picker shows the new model, follow-up completes, persisted thread config has the new model.
Provider switchBuild source history on another provider (include an image turn), then use the thread header's Continue in another provider to switch in place and send a marker.Thread continues on the target provider, prior context transfers, no route/sessionInUse error. The in-thread model picker is provider-locked on active threads — switching goes through the header control.
Reopen/resumeLeave the thread (close the pane), reopen it, and send one more harmless prompt.Existing transcript is not duplicated, session reference remains usable, reply completes. A thread that sits "working" for minutes with no item activity and no error is a hang — FAIL, not slowness.

Do not substitute a successful completion for Stop: a response that ends before the Stop click is not tested. Do not silently claim a provider supports questions, permissions, or model changes when its capabilities did not expose them; report those as not applicable with the advertised capability snapshot.

ACP connection analysis

For a real ACP provider, prove the full path rather than only the rendered reply:

  1. Record availability/authentication and the probed capabilities: models, modes, effort tiers, approval policies, slash commands, and presentation modes.
  2. Confirm createStructuredSession() succeeds and that session/new or session/load yields a stable session reference. For Kimi, verify its discovered providerSessionId is persisted after its delayed session-file discovery.
  3. Confirm session/update messages produce the expected canonical runtime items: user message, reasoning, assistant message, tool call, and tool result/command result when applicable.
  4. Confirm each blocking request receives exactly one request.resolved outcome: answered for a question, declined for a rejected approval, or cancelled for an abandoned request.
  5. For the model gate, prove the update reached the live ACP session config (not merely the picker): inspect the persisted thread config and ensure the next turn uses it.
  6. Reopen the session and verify replayed ACP history does not duplicate Poracode’s persisted chat items.

Capture session identity and configs without copying credentials, bearer tokens, or raw sensitive environment values into artifacts.

Show full SKILL.md (561 more words)Show less

Deterministic ACP branch matrix

Cover every branch below with focused tests/mocks when the live provider does not safely expose it. Add or keep a production scenario mapping for any new provider surface.

AreaBranches to cover
Handshakespawn failure, initialize/protocol mismatch, capability probe timeout, authenticated/missing auth, session/new, and session/load.
Session lifecycleGUI resume versus terminal resume gating, known/discovered session refs, invalid/expired session recovery, process exit/connection loss, dispose, no duplicate history on replay, and failed restart settles the submitted turn (a restart/resume failure — e.g. a sessionInUse-style lock error — must append a visible error item and errored thread state; it must never strand the thread "working" with the send lost).
WSL executionFor providers executing through WSL (windowsProjectExecution: "wsl"): the Linux-side host must not survive teardown — killing the Windows wrapper is not enough, so dispose kills the surviving Linux process (host-cookie environ sweep via the bridge) and no orphan host is left holding provider-side session locks after thread stop or app stop.
Provider routingWhen the provider protocol has internal model/provider routing (e.g. Muse session/start.providerId), launch params must carry the catalog-resolved route rather than the server default — defaults can differ in capability (one Muse route rejects media in retained history, the catalog route accepts it).
Updatesassistant text, reasoning, tool call/update/result, plan/file-change updates, malformed/noise filtering, provider-specific transforms, and empty-response error rewriting.
Config syncmodel, mode, effort/thought-level changes through config options; explicit response versus later notification; unsupported/unstable-model fallback; rejected/timeout config updates.
Promptsnormal completion, RPC error, agent-visible error, cancellation before prompt acceptance, cancellation after activity, pending-steer replacement semantics, Stop watchdog, stale interrupt immunity after restart, and image-attachment delivery per readsImageAttachmentsFromHost (inline image content versus @path mention after execution-location rewrite).
Requestspermission accept/deny/cancel, question options/custom answer/skip (the question form must not submit unanswered questions — the supervisor rejects empty answers), createElicitation, completeElicitation, request resolution after teardown, and synthetic auto-approval only where the configured policy permits it.
Client servicesclient-hosted terminal create/write/output/wait/release, read/write resource path validation, and MCP config/launch gating. Test writes only with the fixture repository.
Provider-route failure signatures

When a live turn fails, classify the error before writing the verdict — some rejections come from the provider's own model routing, not from Poracode:

  • provider-private history is incompatible with the active route: retained media history is unsupported … — the provider's model route refused media in the conversation history. Poracode delivered an image the route cannot take; fix delivery (path mention + readsImageAttachmentsFromHost: false), not the error handling.
  • session … is already in use (sessionInUse) — another live host holds the provider-side session lock, typically an orphaned WSL host that teardown failed to kill. Poracode must surface this as a turn error, never hang.
  • A slow resume of a large session is provider-side while runtime records keep streaming; a hang is "working" with no item activity and no records.

Reporting and teardown

Report a verdict per provider and per gate: PASS, FAIL, SKIPPED, or NOT APPLICABLE. Include connection/capability evidence, session-reference continuity, error count with the first three errors, and screenshot/report paths. Distinguish real provider evidence from mocked protocol coverage.

For WSL-executing providers, teardown hygiene is part of the verdict: after stopping a thread and after stopping the app, check the distro for surviving provider host processes (ps aux | grep <binary>) — a survivor holds provider-side locks and breaks the next resume.

Reset the isolated profile and stop only the process launched for this run. Leave the smoke directory intact unless cleanup is requested.

© Porabuild, Apache-2.0. 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 1 other file in .agents/skills/provider-chat-smoke of Porabuild/Poracode.

  • SKILL.md
  • agents/openai.yaml

Open the folder on GitHubat commit e56fcc8

Compare with similar skills

Provider Chat Smoke 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.

Provider Chat Smoke compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Provider Chat Smoke this skillPorabuild/Poracode115—~4.2kAutomated safety check: PassApache-2.0
Weavebench Cua ReproduceAMAP-ML/LongHorizon-Harness1.7k—~1.6kAutomated safety check: PassMIT
tmux Real User TestingQwenLM/qwen-code28k—~2.3kAutomated safety check: PassApache-2.0
Run Shuntpleaseai/shunt203—~2.6kAutomated safety check: PassApache-2.0
Qwen Mtp Gguf ReleaseR6410418/Jackrong-llm-finetuning-guide1.7k—~474Automated safety check: PassApache-2.0
Qwen Mtp GgufR6410418/Jackrong-llm-finetuning-guide1.7k—~1.7kAutomated safety check: PassMIT

Similar skills

  • Weavebench Cua Reproduce

    AMAP-ML/LongHorizon-Harness

    Reproduce CUA-Harness experiments on WeaveBench from a GitHub checkout.

    1.7k GitHub stars~1.6k tokensUpdated 1 mo ago
    Testing & QAAuto-check passed

  • tmux Real User Testing

    QwenLM/qwen-code

    Drives Qwen Code in a real tmux session the way a user would and saves a readable step-by-step transcript of each screen for maintainers to review.

    28k GitHub stars~2.3k tokensUpdated today
    Testing & QAAuto-check passed

  • Run Shunt

    pleaseai/shunt

    Build, launch, and drive shunt — the Claude Code LLM gateway (a Rust/axum Anthropic-Messages proxy).

    203 GitHub stars~2.6k tokensUpdated yesterday
    Testing & QAAuto-check passed

  • Qwen Mtp Gguf Release

    R6410418/Jackrong-llm-finetuning-guide

    Repository-level wrapper for the canonical Qwen MTP or nextn GGUF release workflow.

    1.7k GitHub stars~474 tokensUpdated 3 mo ago
    Testing & QAAuto-check passed

  • Qwen Mtp Gguf

    R6410418/Jackrong-llm-finetuning-guide

    Complete agent-ready workflow for Qwen-family MTP or nextn GGUF conversion and release.

    1.7k GitHub stars~1.7k tokensUpdated 3 mo ago
    AI & LLM EngineeringAuto-check passed

  • Agenttrace Session Audit

    luoyuctl/agenttrace

    Audit local AI coding-agent sessions with agenttrace. An agent skill from luoyuctl/agenttrace.

    142 GitHub starsUsed in 1 repo~656 tokens
    AI & LLM EngineeringAuto-check passed

More from Porabuild/Poracode

All 21 skills in this repo

  • Interactive Testing

    Porabuild/Poracode

    Run repeatable integration and smoke testing against the real Poracode Electron app through Chrome DevTools Protocol.

    115 GitHub stars~4.5k tokensUpdated today
    Auto-check: notes

  • Release Notes

    Porabuild/Poracode

    Write consistent, hand-written-looking changelog entries for a Poracode release.

    115 GitHub stars~4.2k tokensUpdated today
    Auto-check passed

  • Skill Creator Poracode

    Porabuild/Poracode

    Create a new reusable agent skill managed by Poracode. An agent skill from Porabuild/Poracode.

    115 GitHub stars~1.2k tokensUpdated today
    Auto-check passed

  • App Controls

    Porabuild/Poracode

    Inspect and drive Poracode itself — the Terminal panel, other threads, projects, workspaces, git, pull requests, and schedules — through the poracode MCP.

    115 GitHub stars~745 tokensUpdated today
    Auto-check passed

  • Chrome Control

    Porabuild/Poracode

    Use the user's real Chrome tabs and signed-in sessions for visible browser workflows.

    115 GitHub stars~691 tokensUpdated today
    Auto-check passed

  • Computer Use

    Porabuild/Poracode

    Inspect and operate native Windows, macOS, or Linux applications through Poracode's desktop-control tools.

    115 GitHub stars~718 tokensUpdated today
    Auto-check passed

Works with

Categories

Questions about Provider Chat Smoke

What does Provider Chat Smoke do?

Smoke test real Poracode provider chat threads and ACP sessions end to end. Provider Chat Smoke is an agent skill from Porabuild/Poracode. Smoke test real Poracode provider chat threads and ACP sessions end to end.

When should I use Provider Chat Smoke?

Provider Chat Smoke fits situations like: validating Qwen Code; another structured/ACP provider; testing chat turn handling; permission tools.

How do I install Provider Chat Smoke in Claude Code?

Run `npx skills add Porabuild/Poracode --skill provider-chat-smoke -a claude-code`. Or copy the skill folder (.agents/skills/provider-chat-smoke in Porabuild/Poracode) into .claude/skills/provider-chat-smoke in your project. Claude Code loads it when a task matches its description.

How do I install Provider Chat Smoke in Codex?

Run `npx skills add Porabuild/Poracode --skill provider-chat-smoke -a codex`. Or copy the skill folder (.agents/skills/provider-chat-smoke in Porabuild/Poracode) into .agents/skills/provider-chat-smoke in your project. Codex loads it when a task matches its description.

Can I use Provider Chat Smoke 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 Porabuild/Poracode --skill provider-chat-smoke -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/provider-chat-smoke, .gemini/skills/provider-chat-smoke, .github/skills/provider-chat-smoke and .opencode/skills/provider-chat-smoke in your project.

What does Provider Chat Smoke need to run?

SKILL.md names no scripts, command-line tools or credentials: Provider Chat Smoke is instructions for the agent only.

Does Provider Chat Smoke access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Provider Chat Smoke 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 Provider Chat Smoke use?

Provider Chat Smoke is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Provider Chat Smoke use?

About 4.2k tokens (SKILL.md is roughly 17k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Provider Chat Smoke?

Skills that share tags, products or a category with Provider Chat Smoke: Weavebench Cua Reproduce (AMAP-ML/LongHorizon-Harness, 1.7k stars), tmux Real User Testing (QwenLM/qwen-code, 28k stars), Run Shunt (pleaseai/shunt, 203 stars) and Qwen Mtp Gguf Release (R6410418/Jackrong-llm-finetuning-guide, 1.7k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Provider Chat Smoke?

Porabuild (a GitHub organization) maintains it in Porabuild/Poracode, which has 115 GitHub stars. The repository holds 21 skills in this directory. The repository was last updated on October 10, 2026.

Source: Porabuild/Poracode on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.