# Send Coding Lead to your agent Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually. ## Fast path - Download the package from Yavira. - Extract it into a folder your agent can access. - Paste one of the prompts below and point your agent at the extracted folder. ## Suggested prompts ### New install ```text I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete. ``` ### Upgrade existing ```text I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run. ``` ## Machine-readable fields ```json { "schemaVersion": "1.0", "item": { "slug": "coding-lead", "name": "Coding Lead", "source": "tencent", "type": "skill", "category": "开发工具", "sourceUrl": "https://clawhub.ai/beyound87/coding-lead", "canonicalUrl": "https://clawhub.ai/beyound87/coding-lead", "targetPlatform": "OpenClaw" }, "install": { "downloadUrl": "/downloads/coding-lead", "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=coding-lead", "sourcePlatform": "tencent", "targetPlatform": "OpenClaw", "packageFormat": "ZIP package", "primaryDoc": "SKILL.md", "includedAssets": [ "references/complex-tasks.md", "references/prompt-templates.md", "SKILL.md" ], "downloadMode": "redirect", "sourceHealth": { "source": "tencent", "slug": "coding-lead", "status": "healthy", "reason": "direct_download_ok", "recommendedAction": "download", "checkedAt": "2026-05-01T23:39:34.321Z", "expiresAt": "2026-05-08T23:39:34.321Z", "httpStatus": 200, "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=coding-lead", "contentType": "application/zip", "probeMethod": "head", "details": { "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=coding-lead", "contentDisposition": "attachment; filename=\"coding-lead-3.0.6.zip\"", "redirectLocation": null, "bodySnippet": null, "slug": "coding-lead" }, "scope": "item", "summary": "Item download looks usable.", "detail": "Yavira can redirect you to the upstream package for this item.", "primaryActionLabel": "Download for OpenClaw", "primaryActionHref": "/downloads/coding-lead" }, "validation": { "installChecklist": [ "Use the Yavira download entry.", "Review SKILL.md after the package is downloaded.", "Confirm the extracted package contains the expected setup assets." ], "postInstallChecks": [ "Confirm the extracted package includes the expected docs or setup files.", "Validate the skill or prompts are available in your target agent workspace.", "Capture any manual follow-up steps the agent could not complete." ] } }, "links": { "detailUrl": "https://openagent3.xyz/skills/coding-lead", "downloadUrl": "https://openagent3.xyz/downloads/coding-lead", "agentUrl": "https://openagent3.xyz/skills/coding-lead/agent", "manifestUrl": "https://openagent3.xyz/skills/coding-lead/agent.json", "briefUrl": "https://openagent3.xyz/skills/coding-lead/agent.md" } } ``` ## Documentation ### Coding Lead This skill supersedes inline coding rules in agent SOUL.md files. Route by complexity. ACP fails → auto-fallback to direct execution. ### Task Classification LevelCriteriaActionSimpleSingle file, <60 linesDirect: read/write/edit/execMedium2-5 files, clear scopeACP → fallback directComplexArchitecture, multi-modulePlan → ACP → fallback chunked direct When in doubt, go one level up. ### Tech Stack (New Projects) LayerPreferredFallbackBackendPHP (Laravel/ThinkPHP)PythonFrontendVue.jsReactMobileFlutterUniApp-XCSSTailwind-DBMySQLPostgreSQL Existing projects: follow current stack. New: propose first, wait for confirmation. ### Tool Detection & Fallback All tools are optional. Detect once per session: ToolAvailable?Fallbacksmart-agent-memorynode ~/.openclaw/skills/smart-agent-memory/scripts/memory-cli.js stats ok?memory_search + manual .md writesqmdqmd --version ok?grep (Linux/macOS) / Select-String (Windows) / findACPsessions_spawn succeeds?Direct read/write/edit/exec Notation: [memory] [qmd] [acp] = use if available, fallback if not. ### ACP Agent Routing ACP access: sessions_spawn(runtime="acp") may not be available (platform limitation). Use acpx CLI as the reliable path: # Detect paths dynamically (run once per session) # Windows: # ACPX = %APPDATA%\\npm\\node_modules\\openclaw\\extensions\\acpx\\node_modules\\.bin\\acpx.cmd # AGENT = node %APPDATA%/npm/node_modules/@zed-industries/claude-agent-acp/dist/index.js # macOS/Linux: # ACPX = $(npm root -g)/openclaw/extensions/acpx/node_modules/.bin/acpx # AGENT = node $(npm root -g)/@zed-industries/claude-agent-acp/dist/index.js # One-shot execution via acpx (works from any session type) acpx --verbose --agent "" exec "" 2>&1 If sessions_spawn(runtime="acp") is available in your session, use it. Otherwise fall back to acpx CLI via exec. Check acp.allowedAgents in openclaw.json for available agents: Task TypeBest AgentWhyComplex backend, multi-file refactor, deep reasoningclaudeSuperior cross-file reasoning, long contextQuick iteration, autonomous exploration, sandbox taskscodexFast, autonomous, good at iterative fixesCode reviewDifferent agent than the one that wrote itAvoid same-bias blind spots ### How to route # Default (uses acp.defaultAgent, typically "claude"): sessions_spawn(runtime: "acp", task: , cwd: ) # Explicit agent: sessions_spawn(runtime: "acp", agentId: "claude", task: , cwd: ) sessions_spawn(runtime: "acp", agentId: "codex", task: , cwd: ) ### Fallback chain Try preferred agent → 2. Try alternate agent → 3. Direct execution If one agent fails/unavailable, try the other before falling back to direct execution. ### Parallel with different agents For complex tasks with independent sub-tasks (max 2 parallel): Session 1: claude → backend refactor (needs deep reasoning) Session 2: codex → frontend fixes (needs fast iteration) ### Layer 1: Project-level (Claude Code owns) Projects may have their own CLAUDE.md, .cursorrules, docs/ — these are Claude Code's responsibility. It reads them automatically. Do NOT paste project-level rules into ACP prompts. ### Layer 2: Team-level (OpenClaw owns) shared/knowledge/tech-standards.md — cross-project standards (security, change control, tech stack preferences). Only relevant for direct execution (simple tasks without ACP). ### When spawning ACP Don't embed coding standards in the prompt — Claude Code has its own CLAUDE.md Do include: task description, acceptance criteria, relevant context (file paths, decisions) Do include task-specific constraints if any (e.g., "don't change the API contract") ### When executing directly (no ACP) Load standards once per session, first match wins: shared/knowledge/tech-standards.md (team-level, if exists) Built-in defaults (below, if nothing exists) ### Built-in Defaults (fallback for direct execution) KISS + SOLID + DRY, research before modifying Methods <200 lines, follow existing architecture No hardcoded secrets, minimal change scope, clear commits DB changes via SQL scripts, new tech requires confirmation ### Simple Tasks Read target file(s) (standards already loaded per above) [memory] Recall related decisions Execute with read/write/edit/exec [memory] Record what changed and why ### Step 1: Build Context File Write to /.openclaw/context-.md (ACP reads from disk, not from prompt): # [qmd] or grep: find relevant code # [memory] recall + lessons: find past decisions # Standards already loaded (see "Coding Standards Loading" above) # Write context file with 3-5 key rules from loaded standards — do NOT paste full file Minimal context file structure: # Task Context: ## Project — path, stack, architecture style ## Relevant Code — file paths + brief descriptions from qmd/grep ## History — past decisions/lessons from memory (if any) ## Constraints — task-specific rules only (NOT general coding standards — Claude Code has CLAUDE.md) Full template with examples → see references/prompt-templates.md ### Step 2: Lean Prompt Project: | Stack: Context file: .openclaw/context-.md (read it first if it exists) ## Task ## Acceptance Criteria - [ ] - [ ] Tests pass, no unrelated changes, clean code Before finishing: run linter + tests, include results. When done: openclaw system event --text "Done: " --mode now ### Step 3: Spawn # Option A: sessions_spawn (if available in your session) sessions_spawn(runtime: "acp", task: , cwd: , mode: "run") # Option B: acpx CLI (always works, detect AGENT path per above) exec: acpx --agent "" exec "" # Set cwd to project dir in exec command ### Step 4: Fallback Detection ConditionActionSpawn failed / timeout→ Direct executionEmpty output / no file changes→ Direct executionPartial completion→ Agent fixes remaining Fallback: [memory] log failure → agent executes directly → report to user. Never silently fail. Always complete or report why not. ### Step 5: Verify & Record Check acceptance criteria + run tests [memory] Record: what changed, decisions, lessons Clean up context file ### Complex Tasks Read references/complex-tasks.md only for Complex-level tasks — roles, QA isolation, parallel strategies, RESEARCH→PLAN→EXECUTE→REVIEW flow. ### Context Reuse (Token Savings) Context file on disk instead of prompt embedding → ~90% token savings per spawn Parallel: one context file, multiple ACP sessions read it Serial: use mode: "session" + sessions_send for follow-ups [qmd]: precision search → only relevant snippets in context file No standards in ACP prompts: Claude Code reads its own CLAUDE.md/.cursorrules — don't duplicate ACP prompt stays lean: task + acceptance criteria + context file path. No generic rules Direct execution: load team standards once per session, not per task ### Memory Integration [memory] Before: recall related work + lessons for context file. [memory] After: record what changed, decisions made, lessons learned. Cross-session: agent remembers across sessions; Claude Code doesn't. This is the core advantage. ### Multi-Project Parallel Each project gets its own context file in its own .openclaw/ dir Spawn with different cwd per project — zero cross-contamination Tag memory entries per project: --tags code, Max 2 parallel ACP sessions — keep token/resource use predictable ACP runs in background while agent works on simple tasks directly See references/prompt-templates.md for multi-project examples. ### Smart Retry (max 3) Analyze failure → 2. Adjust prompt → 3. Retry improved → 4. Max 3 then fallback/report. Each retry must be meaningfully different. ### Progress Updates Start → 1 short message. Error → immediate report. Completion → summary. Fallback → explain. ### Safety Never spawn in ~/.openclaw/ — coding agents may damage config Always set cwd to project directory Review before commit — especially complex tasks Kill runaway sessions — timeout or nonsensical output ### See Also references/complex-tasks.md — roles, QA, parallel (Complex only) references/prompt-templates.md — context file template, prompt examples ## Trust - Source: tencent - Verification: Indexed source record - Publisher: beyound87 - Version: 1.0.30 ## Source health - Status: healthy - Item download looks usable. - Yavira can redirect you to the upstream package for this item. - Health scope: item - Reason: direct_download_ok - Checked at: 2026-05-01T23:39:34.321Z - Expires at: 2026-05-08T23:39:34.321Z - Recommended action: Download for OpenClaw ## Links - [Detail page](https://openagent3.xyz/skills/coding-lead) - [Send to Agent page](https://openagent3.xyz/skills/coding-lead/agent) - [JSON manifest](https://openagent3.xyz/skills/coding-lead/agent.json) - [Markdown brief](https://openagent3.xyz/skills/coding-lead/agent.md) - [Download page](https://openagent3.xyz/downloads/coding-lead)