laimis91/assistant-framework/skills/assistant-workflow/SKILL.md
assistant-workflow
Run proportional development phases and resume persisted task state. Use to plan, build, implement, fix, migrate, refactor, or continue project artifacts.
- Source repository stars
- 7
- Declared platforms
- 0
- Static risk flags
- 1
- Last source update
- 2026-08-04
- Source checked
- 2026-08-04
Decision brief
What it does—and where it fits
Public routing contract; detailed mechanics live in references and contracts.
Not for
- Tasks that require unconfirmed production actions or broad system permissions.
- Environments where the pinned source and install steps cannot be inspected.
Compatibility matrix
Platform support, with evidence labels
| Platform | Status | Evidence | What to check |
|---|---|---|---|
| Codex | Not declared | No explicit evidence | Portability before use |
| Claude Code | Not declared | No explicit evidence | Portability before use |
| Cursor | Not declared | No explicit evidence | Portability before use |
| Gemini CLI | Not declared | No explicit evidence | Portability before use |
Installation
Inspect first. Install second.
The source command is displayed only when detected. A safe inspection prompt is always available so your agent can explain every action before execution.
npx skills add https://github.com/laimis91/assistant-framework --skill "skills/assistant-workflow"Inspect the Agent Skill "assistant-workflow" from https://github.com/laimis91/assistant-framework/blob/e85fbadd83b022d9644bbe51b11784cf2e2dfd72/skills/assistant-workflow/SKILL.md at commit e85fbadd83b022d9644bbe51b11784cf2e2dfd72. List every install step, command, network request, credential, file read/write, external action, and rollback step. Explain whether it fits my task. Do not install or execute anything until I approve.
Workflow
What the source asks the agent to do
- 01
Phase Routing
Load references/phases.md for the current phase. Load references/workflow-controller.md only when resolving shared routing/default, movement, harness, review, QA, or subagent-separation decisions. Use references/context-budget-and-pattern-retrieval.md for large material or frame…
Load references/phases.md for the current phase. Load references/workflow-controller.md only when resolving shared routing/default, movement, harness, review, QA, or subagent-separation decisions. Use references/context…For subagent rules, load references/subagent-dispatch.md and resolve subagentpolicystate, subagentexecutionmode, subagenttriggerscope, and conditional policyblockingsource before spawning. Light small low-risk localized…Load references/harness-controller.md only after references/workflow-controller.md or carried-forward phase state establishes harnesscapable=true. Load assistant-security when touching auth, user input, secrets, persist… - 02
Goal
Move work to verified outcome through right-sized phases, gates, tests, review, and company-safe execution.
Move work to verified outcome through right-sized phases, gates, tests, review, and company-safe execution. - 03
Success Criteria
Scale phases to risk; Decompose, Design, and durable state run only when triggered.
Scale phases to risk; Decompose, Design, and durable state run only when triggered.Before resume, reconcile state with the newest request and repository evidence.If resume reconciliation classifies persisted state as stale, superseded, or completed, update the framework-owned {agentstatedir}/task.md before acting or returning; record the classification and reason, current task i… - 04
Constraints
Explicit user or repository artifact schemas override workflow-internal shapes; preserve exact paths, keys, types, ids, and supplied literals.
Explicit user or repository artifact schemas override workflow-internal shapes; preserve exact paths, keys, types, ids, and supplied literals.Do not skip applicable phases; Plan is inapplicable only when the explicit planmode=none eligibility gate passes.Do not ask ritual questions when code/context makes the next safe action clear. - 05
Contracts
Canonical files stay authoritative; validate applicable rules at enforcement points. Read contracts/index.yaml first; do not load every contract at entry. Load only the contract selector applicable to current boundary:
entry: load the exact entry fields declared by contracts/index.yaml from contracts/input.yaml; references/triage-rubric.md is the only declared entry reference.delegation: load role and trigger fields when roles may be required and before any subagent dispatch.currentphase: active contracts/phase-gates.yaml at transition.
Permission review
Static risk signals and limitations
Reads files
The documentation asks the agent to read local files, directories, or repositories.
| Discover | All | Read repo, resolve unknowns, restate requirements. Medium+: Code Mapper map. Unknown-cause bugfixes: load `assistant-debugging`. |Evidence record
Why each signal appears
| Signal | Value | Evidence type | Meaning |
|---|---|---|---|
| Quality score | 92/100 | Computed | Documentation, specificity, maintenance, and trust rules |
| Repository stars | 7 | Source | Repository attention, not individual Skill quality |
| Compatibility | 0 platforms | Source | Declared in the catalog source record |
| Usage guide | automated source guide | Editorial | Generated or reviewed according to the visible evidence level |
Pinned source
Provenance and original SKILL.md
- Repository
- laimis91/assistant-framework
- Skill path
- skills/assistant-workflow/SKILL.md
- Commit
- e85fbadd83b022d9644bbe51b11784cf2e2dfd72
- License
- MIT
- Collected
- 2026-08-04
- Default branch
- main
View the original SKILL.md
Development Workflow
Public routing contract; detailed mechanics live in references and contracts.
Goal
Move work to verified outcome through right-sized phases, gates, tests, review, and company-safe execution.
Success Criteria
- Scale phases to risk; Decompose, Design, and durable state run only when triggered.
- Before resume, reconcile state with the newest request and repository evidence.
- If resume reconciliation classifies persisted state as stale, superseded, or completed, update the framework-owned
{agent_state_dir}/task.mdbefore acting or returning; record the classification and reason, current task identity, and repaired exact next action. plan_modekeeps planning proportional: trivial safe work may usenone, bounded small work may useinline, and medium+, risky, destructive, or scope-shaping work usesapproval_required. Whenplan_mode=inline, small work has an inline plan and proceeds without ceremony unless risk requires approval.references/workflow-controller.mdis the canonical source for controller intensity, workflow state, manual verification, harness/QA routing, and review-role separation.- Ordinary medium+ workflow tasks stay standard, non-harness, and non-QA unless explicit controller criteria apply.
- Harness-capable work carries the Done Contract, Harness Recipe, and trace/replay artifacts required by the controller.
- Candidate Search is reserved for explicit alternatives, open-ended architecture/design, optimization, high uncertainty, repeated failures, unclear/flaky bugs, or reviewer-requested pivots.
- Behavior changes default tests-first or carry explicit validation in the same Build step.
- Review, QA, and security routing apply when triggered.
- Medium+ final output follows
references/final-handoff.md; small output gives changed files, evidence, review status, risks, and next steps.
Constraints
- Explicit user or repository artifact schemas override workflow-internal shapes; preserve exact paths, keys, types, ids, and supplied literals.
- Do not skip applicable phases; Plan is inapplicable only when the explicit
plan_mode=noneeligibility gate passes. - Do not ask ritual questions when code/context makes the next safe action clear.
- Ask bounded clarification before planning only when an undiscoverable implementation-shaping unknown lacks a safe default and affects correctness, scope, behavior, data, public contract, security, migration safety, or verification.
- Keep scope changes explicit and tied to correctness, security, safety, or verification.
- Do not install tools, upload code, call external services, or paste proprietary content into third-party systems unless the user explicitly approved that path.
- Prefer repo-native commands over new tooling.
- When external services, installs, or sensitive-data handling are in scope, load
references/ai-usage-policy.mdfor company-safe detail. It is not an entry dependency.
Contracts
Canonical files stay authoritative; validate applicable rules at enforcement points. Read contracts/index.yaml first; do not load every contract at entry. Load only the contract selector applicable to current boundary:
entry: load the exact entry fields declared bycontracts/index.yamlfromcontracts/input.yaml;references/triage-rubric.mdis the only declared entry reference.delegation: load role and trigger fields when roles may be required and before any subagent dispatch.current_phase: activecontracts/phase-gates.yamlat transition.selected_handoff:contracts/handoffs.yamlbefore dispatch and return validation.completion:contracts/output.yamlat completion before final exit.
Selectors resolve by unique id plus canonical path, exact section, key, and explicit names. Runtime selectors resolve name_from only through their declared allowed_names.
Missing or invalid selector: load_full_authoritative_file; validate the full named canonical file and record recovery.
Migration note: assistant-workflow contracts are v4. Direct-user, applicable
AGENTS.md, or active-skill instructions trigger delegation; record
subagent_trigger_scope and dispatch without a separate permission question.
Explicit opt-out, real unavailability, or an exact policy block uses evidenced
fallback. verification_command is non-empty argv string[]. assistant-review
v3 owns Reviewer/QAEvaluator handoffs and returns final_summary /
qa_evaluation_result; Reviewer reviewed_scope is non-empty.
Visible Checkpoints
Phase markers are required only for controller_intensity=strict, explicit
project policy, or a user request. Light and standard work still follows every
logical phase and gate, but reports progress with concise natural updates
without exact marker ceremony.
When exact markers are required, use this instruction: Use this exact format:
--- PHASE: [name] ---
For steps within a phase:
>> [step description]
For completion:
--- PHASE: [name] COMPLETE ---
These examples are conditional on the exact-marker triggers above.
Refactor Guidance
- Justify it with a concrete risk only: correctness, security, unsafe change surface, ownership, brittle testing, or poor extension seam.
- Tie incidental or scope-expanding refactors to concrete risk instead of vague framing such as generic convention language, style, cleanliness, or generic improvement.
- Choose the smallest useful, durable fix that removes the identified risk. Keep cleanup scoped unless the user explicitly requested cleanup, reorganization, or refactor work.
Triage
Start Triage with a concise progress update. When exact markers are required,
use --- PHASE: TRIAGE ---.
Load references/triage-rubric.md. Perform a quick read-only Candidate scope scan, then assess task type, risk tier, size, gates, agents, controller_intensity, plan_mode, subagent state, and search_mode. For ideas, create binary observable criteria before planning.
| Size | Phases |
|---|---|
| Small | Discover quick -> [Plan] -> Build -> Review -> Document; plan_mode=none only for trivial safe work, otherwise inline or approval-required |
| Medium | Discover -> Decompose -> Plan -> [Design] -> Build -> Review -> Document |
| Large | Discover -> Decompose -> Plan -> Design -> Build -> Review -> Document |
| Mega | Discover -> Decompose -> Plan -> Design -> Build -> Review -> Document |
[Design] = UI only.
Print: >> Triaged as: [SIZE] — phases: [list]
Print: >> Triage metadata: type=[TASK_TYPE] | risk=[RISK_TIER] | intensity=[controller_intensity] | plan=[plan_mode] | gates=[count] | agents=[count] | search=[search_mode] | scope_confidence=[low|medium|high]
If scope exceeds initial triage during any phase, stop and re-triage. Use references/candidate-search.md only when search_mode: candidate_search is selected.
Phase Routing
Load references/phases.md for the current phase. Load references/workflow-controller.md only when resolving shared routing/default, movement, harness, review, QA, or subagent-separation decisions. Use references/context-budget-and-pattern-retrieval.md for large material or framework patterns, references/artifact-first-output-contract.md before Plan, references/decomposition-plan-review.md before medium+ Decompose exits, references/plan-template.md during Plan, and references/context-handoff-templates.md for non-standard continuations.
| Phase | When | Key Actions |
|---|---|---|
| Discover | All | Read repo, resolve unknowns, restate requirements. Medium+: Code Mapper map. Unknown-cause bugfixes: load assistant-debugging. |
| Decompose | Medium+ | Smallest iterable slices with acceptance and verification fields. |
| Plan | plan_mode != none | Inline mode records a compact plan without waiting. Approval-required mode waits for approved scope, slices, packets, files, verification, and risks before Build. |
| Design | UI only | Direction, mockup, checklist, approval gate. |
| Build | All | Light may execute directly; ordinary medium uses one bounded edit/test executor; separated Code Writer -> Builder/Tester is conditional for high-risk or broad/noisy/environment-heavy verification. Tests or relevant validation travel with code. |
| Review | All | Light gets a fresh self-review without worker/reviewer dispatch evidence; standard/strict use Spec Review then independent assistant-review; QA only when required. |
| Document | All | Apply the state and manual-verification modes; metrics remain optional and non-blocking. |
For subagent rules, load references/subagent-dispatch.md and resolve subagent_policy_state, subagent_execution_mode, subagent_trigger_scope, and conditional policy_blocking_source before spawning. Light small low-risk localized work selects not_required plus not_applicable. For standard/strict development work, a direct user request or applicable AGENTS.md or active-skill instruction triggers delegation: set delegation_triggered, infer its covered roles/phases/actions, and dispatch the configured role agents without a separate permission question. Use direct fallback only for explicit user opt-out, an exact active policy block, or real subagent unavailability; do not infer unavailability merely because no visible tool is named Task, delegate, or subagent. Delegation never bypasses parent sandbox, action/tool approvals, external-write, install, destructive-operation, or secrets safeguards.
Load references/harness-controller.md only after references/workflow-controller.md or carried-forward phase state establishes harness_capable=true.
Load assistant-security when touching auth, user input, secrets, persistence, network calls, shell commands, dependency/config changes, or external integrations.
For review-gated multi-slice work, load references/slice-review-topology.md before emitting or consuming slice review evidence.
Output
Return:
- Status: complete, partially complete, blocked, or plan ready.
- Changed files: paths and purpose.
- Verification: commands, pass/fail, skipped checks with reasons.
- Review result: spec, quality, QA, and security when applicable.
- Residual risk: blockers, assumptions, policy constraints, or follow-up.
- Next step: one practical recommendation.
Do not use phrases like "should work", "probably fixed", or "looks good" unless immediately qualified with evidence or uncertainty.
Stop Rules
- Stop and ask when an implementation-shaping field is material, undiscoverable, and has no safe default.
- Stop before Build when
plan_mode=approval_requireduntil plan approval. - Stop before final response if build, tests, required review, or output contract evidence is missing.
- Stop when a plan deviation changes approved scope, files, behavior, risk, verification, or acceptance criteria; record the deviation and get approval before continuing.
Alternatives
Compare before choosing
laimis91/assistant-framework
assistant-workflow
Run proportional development phases and resume persisted task state. Use to plan, build, implement, fix, migrate, refactor, or continue project artifacts.
alirezarezvani/claude-skills
app-store-optimization
App Store Optimization (ASO) toolkit for researching keywords, analyzing competitor rankings, generating metadata suggestions, and improving app visibility on Apple App Store and Google Play Store. Use when the user asks about ASO, app store rankings, app metadata, app titles and descriptions, app store listings, app visibility, or mobile app marketing on iOS or Android. Supports keyword research and scoring, competitor keyword analysis, metadata optimization, A/B test planning, launch checklist
wanshuiyin/Auto-claude-code-research-in-sleep
citation-audit
Use it for operations and research tasks; the detail page covers purpose, installation, and practical steps.
dotnet/skills
migrate-vstest-to-mtp
Migrates .NET test projects from VSTest to Microsoft.Testing.Platform (MTP). Use when user asks to "migrate to MTP", "switch from VSTest", "enable Microsoft.Testing.Platform", "use MTP runner", set OutputType=Exe only for test projects in Directory.Build.props, or mentions EnableMSTestRunner, EnableNUnitRunner, or UseMicrosoftTestingPlatformRunner. USE FOR: MTP behavioral differences vs VSTest (exit code 8, zero tests discovered, --ignore-exit-code, TESTINGPLATFORM_EXITCODE_IGNORE); centralizing