{"id":"bcc08066-02bc-4125-a069-dd8c9149086c","shortId":"BLPzFe","kind":"skill","title":"work","tagline":"This skill should be run only when the user explicitly invokes it. Orchestrates end-to-end task implementation — understands the task, assesses complexity, implements directly or via a team of subagents for complex work, and always finishes with a code-polish pass.","description":"# Work\n\nOrchestrate end-to-end task implementation: understand the task, assess complexity, implement directly or distribute across a team, then polish the result.\n\n## Workflow\n\n### 1) Parse Task\n\nRead the task description from `$ARGUMENTS`.\n\n- If `$ARGUMENTS` is empty, ask the user for a task description and stop.\n- Extract key signals: scope (files, modules, components mentioned), action type (new feature, bug fix, refactor, migration), and any constraints.\n- Note any referenced issues, PRs, or URLs for later context gathering.\n\n### 2) Assess Complexity\n\nClassify the task as **simple** or **complex** using these heuristics:\n\n| Signal           | Simple                   | Complex                                       |\n| ---------------- | ------------------------ | --------------------------------------------- |\n| File count       | 1-3 files                | 4+ files                                      |\n| Module span      | Single module or package | Cross-module or cross-package                 |\n| Dependency chain | No new dependencies      | New packages or service integrations          |\n| Risk surface     | Low (UI, docs, config)   | High (auth, payments, data, infra)            |\n| Parallelism      | Sequential steps only    | Independent subtasks benefit from concurrency |\n\nA task is complex when **3 or more** signals fall in the complex column. When in doubt, prefer the simple path — team overhead is only justified when parallelism provides a real speedup.\n\n- **Simple** — proceed to Step 3.\n- **Complex** — proceed to Step 4.\n\n### 3) Implement (Simple Path)\n\nExecute the task directly without spawning subagents.\n\n1. **Gather context**: Read all relevant files. Understand existing code, tests, and conventions.\n2. **Implement**: Make the changes. Follow project conventions inferred from existing code, linters, and formatters.\n3. **Verify**: Run the narrowest useful checks:\n   - Formatter/linter on touched files.\n   - Targeted tests for touched modules.\n   - Type check when relevant.\n   - If fast checks pass, run broader checks only when risk warrants it.\n4. Proceed to Step 5 (Polish).\n\n### 4) Implement (Complex Path)\n\nDistribute work across a team of subagents.\n\n#### 4a) Decompose\n\nBreak the task into independent subtasks. Each subtask should:\n\n- Target a distinct set of files with minimal overlap.\n- Be completable without waiting on other subtasks (no circular dependencies).\n- Include clear acceptance criteria.\n\nAvoid over-decomposition. If subtasks cannot run in parallel, prefer the simple path.\n\n#### 4b) Create Team and Assign\n\nCreate a team with a name derived from the task (e.g., \"add-auth\", \"refactor-api\"). Create a task for each subtask. Set up dependencies when ordering matters.\n\nSpawn implementation agents as teammates. Assign each agent one or more tasks.\n\nRecommended team sizing:\n\n- 1 agent per module when modules are independent.\n- Separate agent for shared utilities when consumers depend on them (utility blocks consumers).\n- Dedicated agent for tests if test volume is high.\n\n#### 4c) Coordinate\n\nMonitor progress. As agents complete tasks:\n\n- Review output for integration issues.\n- Resolve cross-agent conflicts (merge overlaps, API mismatches).\n- Assign follow-up tasks if gaps emerge.\n\nAfter all tasks complete:\n\n- Run integration verification: full test suite, type check, lint.\n- Fix any integration issues directly — do not re-spawn agents for small fixes.\n- Shut down teammates.\n- Proceed to Step 5 (Polish).\n\n### 5) Polish\n\nInvoke `/code-polish` to simplify and review all session-modified files.\n\nWait for completion. If it reports residual risks or stop conditions, relay them to the user.\n\nThis step is mandatory — always run it, even if the implementation seems clean.\n\n## Error Handling\n\n| Error                                | Response                                                      |\n| ------------------------------------ | ------------------------------------------------------------- |\n| Empty `$ARGUMENTS`                   | Ask for a task description and stop                           |\n| Verification failures after impl     | Attempt to fix; if unfixable, report to user before polishing |\n| Team agent fails or times out        | Reclaim the task and complete it directly                     |\n| `code-polish` reports stop condition | Relay to user with context                                    |\n\n## Stop Conditions\n\nStop and ask for direction when:\n\n- The task description is ambiguous and multiple interpretations exist.\n- Implementation requires changing public APIs or breaking contracts not mentioned in the task.\n- The task scope grows beyond the original description during implementation.\n- External dependencies (APIs, services, packages) are unavailable or broken.\n- A CRITICAL security concern is discovered in existing code adjacent to the task.","tags":["work","agent","skills","paulrberg","agent-skills","ai-agents"],"capabilities":["skill","source-paulrberg","skill-work","topic-agent-skills","topic-ai-agents"],"categories":["agent-skills"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/PaulRBerg/agent-skills/work","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add PaulRBerg/agent-skills","source_repo":"https://github.com/PaulRBerg/agent-skills","install_from":"skills.sh"}},"qualityScore":"0.475","qualityRationale":"deterministic score 0.47 from registry signals: · indexed on github topic:agent-skills · 50 github stars · SKILL.md body (4,870 chars)","verified":false,"liveness":"unknown","lastLivenessCheck":null,"agentReviews":{"count":0,"score_avg":null,"cost_usd_avg":null,"success_rate":null,"latency_p50_ms":null,"narrative_summary":null,"summary_updated_at":null},"enrichmentModel":"deterministic:skill-github:v1","enrichmentVersion":1,"enrichedAt":"2026-04-22T00:56:19.010Z","embedding":null,"createdAt":"2026-04-18T22:17:49.712Z","updatedAt":"2026-04-22T00:56:19.010Z","lastSeenAt":"2026-04-22T00:56:19.010Z","tsv":"'-3':142 '/code-polish':514 '1':71,141,242,416 '2':123,255 '3':194,225,231,270 '4':144,230,302,308 '4a':319 '4b':367 '4c':446 '5':306,509,511 'accept':351 'across':63,314 'action':101 'add':384 'add-auth':383 'adjac':662 'agent':403,408,417,425,438,451,462,499,581 'alway':38,544 'ambigu':616 'api':388,466,625,646 'argument':79,81,558 'ask':84,559,608 'assess':24,57,124 'assign':371,406,468 'attempt':570 'auth':176,385 'avoid':353 'benefit':186 'beyond':638 'block':435 'break':321,627 'broader':295 'broken':652 'bug':105 'cannot':359 'chain':160 'chang':259,623 'check':276,287,292,296,487 'circular':347 'classifi':126 'clean':552 'clear':350 'code':43,251,266,594,661 'code-polish':42,593 'column':202 'complet':340,452,479,526,590 'complex':25,35,58,125,132,138,192,201,226,310 'compon':99 'concern':656 'concurr':188 'condit':534,598,605 'config':174 'conflict':463 'constraint':111 'consum':430,436 'context':121,244,603 'contract':628 'convent':254,262 'coordin':447 'count':140 'creat':368,372,389 'criteria':352 'critic':654 'cross':153,157,461 'cross-ag':460 'cross-modul':152 'cross-packag':156 'data':178 'decompos':320 'decomposit':356 'dedic':437 'depend':159,163,348,397,431,645 'deriv':378 'descript':77,90,563,614,641 'direct':27,60,238,493,592,610 'discov':658 'distinct':332 'distribut':62,312 'doc':173 'doubt':205 'e.g':382 'emerg':475 'empti':83,557 'end':16,18,49,51 'end-to-end':15,48 'error':553,555 'even':547 'execut':235 'exist':250,265,620,660 'explicit':11 'extern':644 'extract':93 'fail':582 'failur':567 'fall':198 'fast':291 'featur':104 'file':97,139,143,145,248,280,335,523 'finish':39 'fix':106,489,502,572 'follow':260,470 'follow-up':469 'formatt':269 'formatter/linter':277 'full':483 'gap':474 'gather':122,243 'grow':637 'handl':554 'heurist':135 'high':175,445 'impl':569 'implement':20,26,53,59,232,256,309,402,550,621,643 'includ':349 'independ':184,325,423 'infer':263 'infra':179 'integr':168,457,481,491 'interpret':619 'invok':12,513 'issu':115,458,492 'justifi':214 'key':94 'later':120 'lint':488 'linter':267 'low':171 'make':257 'mandatori':543 'matter':400 'mention':100,630 'merg':464 'migrat':108 'minim':337 'mismatch':467 'modifi':522 'modul':98,146,149,154,285,419,421 'monitor':448 'multipl':618 'name':377 'narrowest':274 'new':103,162,164 'note':112 'one':409 'orchestr':14,47 'order':399 'origin':640 'output':455 'over-decomposit':354 'overhead':211 'overlap':338,465 'packag':151,158,165,648 'parallel':180,216,362 'pars':72 'pass':45,293 'path':209,234,311,366 'payment':177 'per':418 'polish':44,67,307,510,512,579,595 'prefer':206,363 'proceed':222,227,303,506 'progress':449 'project':261 'provid':217 'prs':116 'public':624 're':497 're-spawn':496 'read':74,245 'real':219 'reclaim':586 'recommend':413 'refactor':107,387 'refactor-api':386 'referenc':114 'relay':535,599 'relev':247,289 'report':529,575,596 'requir':622 'residu':530 'resolv':459 'respons':556 'result':69 'review':454,518 'risk':169,299,531 'run':6,272,294,360,480,545 'scope':96,636 'secur':655 'seem':551 'separ':424 'sequenti':181 'servic':167,647 'session':521 'session-modifi':520 'set':333,395 'share':427 'shut':503 'signal':95,136,197 'simpl':130,137,208,221,233,365 'simplifi':516 'singl':148 'size':415 'skill':3 'skill-work' 'small':501 'source-paulrberg' 'span':147 'spawn':240,401,498 'speedup':220 'step':182,224,229,305,508,541 'stop':92,533,565,597,604,606 'subag':33,241,318 'subtask':185,326,328,345,358,394 'suit':485 'surfac':170 'target':281,330 'task':19,23,52,56,73,76,89,128,190,237,323,381,391,412,453,472,478,562,588,613,633,635,665 'team':31,65,210,316,369,374,414,580 'teammat':405,505 'test':252,282,440,442,484 'time':584 'topic-agent-skills' 'topic-ai-agents' 'touch':279,284 'type':102,286,486 'ui':172 'unavail':650 'understand':21,54,249 'unfix':574 'url':118 'use':133,275 'user':10,86,539,577,601 'util':428,434 'verif':482,566 'verifi':271 'via':29 'volum':443 'wait':342,524 'warrant':300 'without':239,341 'work':1,36,46,313 'workflow':70","prices":[{"id":"aa466f7b-c45c-4e3a-a787-4f4277d1c301","listingId":"bcc08066-02bc-4125-a069-dd8c9149086c","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"PaulRBerg","category":"agent-skills","install_from":"skills.sh"},"createdAt":"2026-04-18T22:17:49.712Z"}],"sources":[{"listingId":"bcc08066-02bc-4125-a069-dd8c9149086c","source":"github","sourceId":"PaulRBerg/agent-skills/work","sourceUrl":"https://github.com/PaulRBerg/agent-skills/tree/main/skills/work","isPrimary":false,"firstSeenAt":"2026-04-18T22:17:49.712Z","lastSeenAt":"2026-04-22T00:56:19.010Z"}],"details":{"listingId":"bcc08066-02bc-4125-a069-dd8c9149086c","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"PaulRBerg","slug":"work","github":{"repo":"PaulRBerg/agent-skills","stars":50,"topics":["agent-skills","ai-agents"],"license":"mit","html_url":"https://github.com/PaulRBerg/agent-skills","pushed_at":"2026-04-20T16:22:56Z","description":"PRB's collection of agent skills","skill_md_sha":"2af35354112a1f8e35dbf61abfa6e922b776e76a","skill_md_path":"skills/work/SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/PaulRBerg/agent-skills/tree/main/skills/work"},"layout":"multi","source":"github","category":"agent-skills","frontmatter":{"name":"work","description":"This skill should be run only when the user explicitly invokes it. Orchestrates end-to-end task implementation — understands the task, assesses complexity, implements directly or via a team of subagents for complex work, and always finishes with a code-polish pass."},"skills_sh_url":"https://skills.sh/PaulRBerg/agent-skills/work"},"updatedAt":"2026-04-22T00:56:19.010Z"}}