@hobin/developer 0.1.0

package/skills/specify/SKILL.md

@@ -0,0 +1,67 @@
+---
+name: specify
+description: "Clarify product meaning, scope, invariants, constraints, counterexamples, and blocking unknowns from a requirement, bug report, code behavior, or implementation request. Use when implementation could be mechanically correct while still being wrong about what must be true."
+---
+
+# Specify
+
+Clarify what must be true without deciding how to build it.
+
+## Core Question
+
+What must be true, in scope, and not guessed?
+
+## Inputs
+
+- Requirement, product request, bug report, or policy
+- Observed code or UI behavior
+- Existing tests, examples, and accepted decisions
+- Constraints, non-goals, and affected users when known
+
+## Output
+
+Lead with the user's product language; keep specification labels secondary.
+Produce product meaning, invariant, scope and non-scope, constraints,
+counterexamples, and blocking unknowns. When used inside a larger task, return:
+
+```text
+Status: resolved | needs-evidence | not-applicable | blocked
+Result: the clarified product contract
+Basis: request, behavior, examples, and explicit assumptions
+Open questions: unresolved product decisions, or none
+Artifacts: invariant, scope, constraints, and counterexamples
+```
+
+Return only this skill's judgment for the question at hand; leave subsequent
+routing to the caller.
+
+## Completion
+
+Finish when the product contract is precise enough that different reasonable
+implementations would preserve the same meaning. Revisit when the user changes
+scope or new behavior contradicts the invariant.
+
+## Method
+
+1. Separate desired outcome from proposed implementation.
+2. State the product meaning in the user's language.
+3. Mark in-scope, out-of-scope, and intentionally unchanged behavior.
+4. State the invariant and turn it into checkable constraints.
+5. Pair each important constraint with a violating counterexample.
+6. Distinguish evidence-backed facts, provisional assumptions, and decisions
+   only the user or product owner can make.
+7. Ask the smallest clarification question only when guessing would materially
+   change the contract.
+
+## Missing Evidence
+
+Return `needs-evidence` when repository or product evidence can answer the
+question. Return `blocked` only when a human-owned product decision is required.
+Return `not-applicable` when product meaning is already accepted and no scope or
+invariant question remains. Bound non-consequential uncertainty with an
+explicit provisional assumption.
+
+## Boundary
+
+Do not build a full condition model, design the implementation, schedule
+structural work, mutate artifacts, or verify completion.

package/skills/verify/SKILL.md

@@ -0,0 +1,71 @@
+---
+name: verify
+description: "Judge what claims are supported by code, tests, diffs, command results, UI states, implementation reports, invariants, or problem models. Use when completion, correctness, coverage, or pass-but-wrong risk depends on mapping current evidence to a concrete claim."
+---
+
+# Verify
+
+Map evidence to claims without turning green checks into broader proof.
+
+## Core Question
+
+What claims does the available evidence support?
+
+## Inputs
+
+- Claims under review
+- Test or command output, diff, code, UI state, or implementation report
+- Invariant, model, accepted decisions, and verification targets when available
+- Relevant counterexamples and source provenance
+
+## Output
+
+Lead with the user's claim and observed evidence; keep verification labels secondary.
+Produce a scoped satisfaction judgment, evidence map, target and counterexample
+coverage, evidence strength, wrong implementations that could still pass,
+assumptions, unverified claims, and residual risk. When used inside a larger
+task, return:
+
+```text
+Status: resolved | needs-evidence | not-applicable | blocked
+Result: the strongest claim justified by the evidence
+Basis: checks, inspections, provenance, and accepted assumptions
+Open questions: unverified claims or human-owned acceptance, or none
+Artifacts: evidence map and residual risks
+```
+
+Return only this skill's judgment for the question at hand; leave subsequent
+routing to the caller.
+
+## Completion
+
+Finish when each stated claim is supported, narrowed, or explicitly left
+unverified, and plausible pass-but-wrong shapes have been considered. Revisit
+after code changes, new evidence, changed claims, or source-provenance changes.
+
+## Method
+
+1. Extract the exact claims from the request, accepted decisions, implementation
+   report, invariant, or model.
+2. Attach concrete evidence to each claim and classify its strength.
+3. Check constraints, forbidden cases, transitions, callers, and abstraction
+   stop checks when they are part of the claim.
+4. Distinguish verifier execution from verifier relevance.
+5. Ask what wrong implementation could still pass and name its concrete shape.
+6. Narrow any passing claim that is broader than its evidence.
+7. Separate missing evidence, blocked evidence, human judgment, and residual
+   risk instead of hiding them in a passing summary.
+8. For compatibility, bind evidence to the actual source and version. A source
+   mismatch is not evidence about the requested target.
+
+## Missing Evidence
+
+Return `needs-evidence` when an available check or inspection can answer the
+claim. Return `not-applicable` when there is no claim to judge. Return `blocked`
+when required evidence depends on unavailable environment or human acceptance.
+If no claim is stated, infer only the weakest claim justified by the evidence.
+
+## Boundary
+
+Do not invent product scope, run an entire development workflow, repair the
+implementation, or turn weak evidence into a completion claim.

package/skills/visualize/SKILL.md

@@ -0,0 +1,65 @@
+---
+name: visualize
+description: "Choose and shape the smallest visual surface that materially lowers the cost of inspecting a decision, model, evidence set, architecture relation, UI state, quantitative comparison, or workflow. Use when relationships, cases, sequence, hierarchy, or repeated mappings are materially harder to judge in prose."
+---
+
+# Visualize
+
+Make a judgment easier to inspect without making weak evidence look certain.
+
+## Core Question
+
+What visual surface would materially lower judgment cost?
+
+## Inputs
+
+- Decision or inspection question
+- Source facts, model, evidence, or relationships
+- Audience and medium constraints
+- Known uncertainty and interpretations to avoid
+
+## Output
+
+Use the user's concepts as labels; keep visualization terminology secondary.
+Produce whether a visual is useful, the inspection question, recommended form,
+visual content, what it reveals, and what it must not imply. When used inside a
+larger task, return:
+
+```text
+Status: resolved | needs-evidence | not-applicable | blocked
+Result: recommended visual form or a decision to use prose
+Basis: structure of the decision and source evidence
+Open questions: missing facts or uncertain relationships, or none
+Artifacts: visual specification or completed visual
+```
+
+Return only this skill's judgment for the question at hand; leave subsequent
+routing to the caller.
+
+## Completion
+
+Finish when the chosen surface answers the inspection question with less effort
+than prose and uncertainty remains visible. Revisit when the underlying model,
+evidence, or audience changes.
+
+## Method
+
+1. Name the decision the user needs to inspect.
+2. Decline a visual when prose is smaller and clearer.
+3. Match structure to form: table for finite mappings, state machine for allowed
+   movement, timeline for order, relation map for dependencies, evidence matrix
+   for claims and checks, tree for hierarchy, and layered diagram for boundaries.
+4. Use the user's concepts as labels.
+5. Preserve source semantics and mark inferred relationships.
+6. State the overinterpretation the visual must not encourage.
+
+## Missing Evidence
+
+Return `needs-evidence` when missing facts determine the visual structure.
+Return `not-applicable` when prose is the better surface. Return `blocked` when
+the requested medium or source artifact is unavailable.
+
+## Boundary
+
+Do not invent architecture, evidence, product meaning, or certainty. Do not use
+a visual to decorate the workflow or replace the judgment it should support.