dev-companion
Skill19 starsChanged 2 months agoMITView on GitHub
WHAT — Dev Companion (general): layered companion for client delivery; modes, gates, delegation
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 4 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
If engagement triggers match, load the appropriate workspace pack overlay first, then proceed with workflow-generic-project.
- Distinctive: line 27
If the user mentions an engagement, ticket prefix, or repo context → load the matching pack from the workspace (overlay) and apply its …
- Distinctive: line 28
If unclear → ask before applying L3.
- Distinctive: line 53
Do not paste forge or ticket CLI sequences here.
- Very specific: line 58
Queued job (optional): only when a local runner is configured.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Dev Companion (WHAT) — general layer (L2)
This skill is the general dev companion for work. It does not replace assistant (orchestrator); it sits above workflows and names what to invoke next.
Language: English for ticket comments, PR text, and user-facing outputs when this companion drives delivery work.
Layering
| Layer | Skill | Role |
| L1 | assistant | Repo inspection order, conflict resolution, fallback |
| L2 (this skill) | dev-companion | Companion framing: modes, gates, delegation for client work |
| L3 | Workspace pack overlay | Client/account-specific context loaded from ~/.ai-workspace/packs/ |
If engagement triggers match, load the appropriate workspace pack overlay first, then proceed with workflow-generic-project. Do not mix multiple workflow drivers on the same task.
Mode selection
- Default for /client delivery: use workflow-generic-project for phased delivery.
- If the user mentions an engagement, ticket prefix, or repo context → load the matching pack from the workspace (overlay) and apply its boundaries/gates.
- If unclear → ask before applying L3.
Delegation (HOW is in other skills)
| Need | Delegate to |
| Discovery, AGENTS conflict handling | assistant |
| Generic client delivery phases | workflow-generic-project |
| Planning / default workflow / validation | planning / development-workflow |
| Work items, incidents, meetings, decisions, agreements, spikes | work-item, incident, meeting-minutes, decision-log, agreement, spike |
| Project assessments, evidence maps, technical/management unit scorecards | project-assessment, project-assessment-evidence, technical-unit-assessment, management-unit-assessment |
| ClickUp | clickup-cli |
| Jira | External jira-* pack (jira-assistant router; CLI: jira-as) |
| Confluence | External confluence-* pack (confluence-assistant router; CLI: confluence-as or confluence) |
| Draft PR GitHub / GitLab | github-cli-workflow / gitlab-cli-workflow (for default body when the repo has no template: pr-fallback first) |
| CI failures (GitHub Actions) | gh-fix-ci |
| Merge / rebase conflicts | fix-merge-conflicts |
| Prose quality before PR | unslop |
| Code slop in diff | deslop |
| Large change scope | blast-radius |
| Session learnings | workspace-knowledge-sync |
| Where to save deliverables + human review | output-handshake |
| UI depth | figma / figma-implement-design |
| Workstation health | triage |
Do not paste forge or ticket CLI sequences here.
Operating modes
- Interactive (default): IDE session; user steers each step.
- Queued job (optional): only when a local runner is configured; see
~/.local/share//dev-companion/README.md(installed from chezmoi) and references/LOOP_GUARDRAILS.md.
The queue worker is mandatory infrastructure (installed by workstation). The workspace (~/.ai-workspace) is optional — it provides project-aware wrappers, job templates, and knowledge base integration. When both are present, the runner automatically enriches LLM prompts with workspace context (projects.yaml, projects/, knowledge/todos/).
LLM policy gate (client engagements)
Before queueing background jobs for a client repo, confirm the active LLM policy with devcompanion llm-status. If the engagement requires a single AI account (e.g. only the customer's Anthropic key), the workstation/pack must set DOTS_WORKSTATION_DEVCOMPANION_LLM_ALLOWLIST and DOTS_WORKSTATION_DEVCOMPANION_LLM_STRICT=1 so the runner fails closed instead of falling back to OpenCode. For Cursor/Copilot-only engagements, prefer run-once --no-llm (skeleton plan) plus IDE-driven execution. Full reference: docs/DEV_COMPANION_LLM.md.
Checklist
- L3 not needed; if needed, user confirmed engagement context
- assistant discovery pass before large edits
- workflow-generic-project phases and gates when doing generic client delivery
- Tool skills used for all CLI operations
In this repository106 other files it gives its agents.
- AGENTS.mdAGENTS.md
- reviewSkill
- mcp-auditSkill
- owasp-agentic-reviewSkill
- supply-chain-auditSkill
- threat-modelingSkill
- architecture-diagramSkill
- c4-modelSkill
- aws-well-architected-reviewSkill
- cloud-design-patternsSkill
- assistantSkill
- onboardingSkill
- output-handshakeSkill
- pr-fallbackSkill
- projectSkill
- workspace-knowledge-syncSkill
- workspaceSkill
- dbt-validationSkill
- snowflake-validationSkill
- adrSkill
- agreementSkill
- bugSkill
- decision-logSkill
- development-workflowSkill
- epicSkill
- incidentSkill
- management-unit-assessmentSkill
- meeting-minutesSkill
- planningSkill
- prdSkill
- project-assessment-evidenceSkill
- project-assessmentSkill
- spikeSkill
- taskSkill
- technical-unit-assessmentSkill
- trdSkill
- user-storySkill
- workflow-client-bootstrapSkill
- workflow-generic-projectSkill
- work-itemSkill
- bragSkill
- brag-slimSkill
- design-assessmentSkill
- design-improvementSkill
- figma-code-connect-componentsSkill
- figma-create-design-system-rulesSkill
- figma-create-new-fileSkill
- figma-implement-designSkill
- figmaSkill
- frontend-design-reviewSkill
- frontend-designSkill
- scroll-worldSkill
- web-design-guidelinesSkill
- fix-merge-conflictsSkill
- gh-address-commentsSkill
- gh-contribution-plannerSkill
- gh-fix-ciSkill
- github-cli-workflowSkill
- gitlab-cli-workflowSkill
- worktreeSkill
- The first 60 shown.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.