| name | comment-dev |
| description | The front door for engineering work through a Comment.io worklog. Talk to it about dev work in plain language and it picks the right path: shape a rough idea (`comment-spec`), build a defined feature (`comment-feature`), fix a defect (`comment-bug`), or try a fast change you'll validate later (`comment-prototype`). Invoke as `$comment-dev` / `/comment-dev`, or when someone describes coding work — "build / add / implement", "fix / it's broken", "let me try / quick tweak / show me", "should we / scope this" — without naming a specific path. When the user already named a specific skill, let that one fire directly. Works identically under Codex and Claude Code. |
comment-dev — the engineering front door
One thing to talk to. Describe the work; comment-dev classifies the intent and
routes to the path that fits, so you don't have to remember which comment-*
skill to call. It is a thin dispatcher — it does no delivery itself; it picks
a path and hands off.
Not a Comment.io document skill. "Make me a comm / read this doc / edit this
comm" is the comment skill. comment-dev is for building, fixing, and
shaping software.
Routing
Classify the request, state which path you're taking and why (one line), then run
that skill (read its full SKILL.md first — naming it here does not load it):
| The request looks like… | Route to | Why |
|---|
| A rough/unshaped idea; goal, scope, or product fit unsettled; "should we…", "what if…", "scope this", "is this worth doing" | comment-spec | Shape it into a crisp spec first; it then invokes comment-feature. |
| A defined feature; "build / add / implement / wire up X", clear what success is | comment-feature | Full plan → build → review → ship through a worklog. |
| A defect/regression; "X is broken", "this throws", "stopped working", "wrong output" | comment-bug | Reproduce → failing test → fix → verify. |
| A fast, small, or exploratory change to look at first; "let me try", "quick tweak", "just show me", "spike / prototype this", UI nudges before committing to the gate | comment-prototype | Implement fast, show it, skip the heavy gate; promote later. |
| Make / read / edit a Comment.io doc itself (not code) | comment | That's a document operation, not a delivery flow. |
Picking between adjacent paths
- prototype vs feature/bug: if the user wants to see it before investing,
or it's a handful of UI/copy tweaks, start with
comment-prototype — it
promotes into comment-feature/comment-bug when they like it. If they want
it real and merge-ready from the start, go straight to feature/bug.
- spec vs feature: if you can't yet state the goal and what "done" means in
one sentence, it's a spec. If you can, it's a feature.
- feature vs bug: new capability → feature; restoring intended behavior →
bug.
When it's genuinely ambiguous, ask one crisp question with your recommended
route — don't guess on a high-cost path. When it's clear, just route and say so.
What carries across the handoff
- Identity / worklog. The chosen skill opens the worklog and (via
comment-identity) the session-scoped ephemeral handle. If the user already
has a worklog / Project Root URL for this task, pass it through so the routed
skill reuses it instead of opening a second root.
- Repo config. The routed skills read this repo's setup themselves (see
Repo config below);
comment-dev doesn't need to.
Repo config
comment-dev only routes, so it needs no repo specifics. The skills it routes to
read AGENTS.md (else CLAUDE.md) and its linked docs/TESTING.md
(test lanes), PR/deploy docs, and architecture. If those are missing, the routed
skill infers them or offers to run comment-init to scaffold them.
Out of scope
comment-dev never implements, tests, or ships directly — it delegates. If a
request spans paths (e.g. "spec then build"), route to the first one and let
the chain continue (comment-spec → comment-feature).
