Pick For The Problem

Pick for the problem

A drift corrector for selection discipline: the tool must fit the problem, not the reflex. The method — re-anchor, audit the work in flight, correct forward, report, and the tone that firing this is not an accusation — lives in [${CLAUDE_PLUGIN_ROOT}/context/re-anchor-audit-correct.md](../../context/re-anchor-audit-correct.md). Read it; this file adds only what is specific to selecting a tool, library, framework, language, or approach.

The discipline this re-anchors

A selection is a design decision, not a reflex. Resolve the source of truth per the method doc's ladder: if the consuming project states a technology-selection or dependency-adoption rule in its own CLAUDE.md / .claude/rules/, re-anchor THAT. Otherwise re-anchor this portable baseline.

The four selection sins — an unexamined choice usually traces to one:

• Habit — "I always use X." The reach is muscle memory, not analysis.
• Availability — "X is already at hand." Convenience picked it, not fit.
• Incumbency — "the repo already uses X, so new work assumes X." Current

state is treated as the requirement.

• Preconception — "I came in believing X is the answer." The verdict

preceded the problem.

The discipline that replaces them:

• Define the actual problem first. Name what is being solved — the real

requirements — before any candidate is on the table. Do not let the first solution shape decide the problem.

• Survey the field. More than one candidate, judged against the stated

requirements and the plausible future ones (variables that could shift and turn a choice into future pain).

• Walk the preference ladder — an earlier rung wins when it covers the

requirements and the plausible future requirements:

1. Native — what the platform, language, or framework already

provides: no new dependency, no new coupling.

1. Official / authoritative — the first-party or canonical option when

native falls short.

1. Vetted third-party — only when the rungs above genuinely miss, and

only if it is well-maintained, well-known, safe, and secure.

• Every dependency is a coupling point. Weigh, at adoption time, the

cost this coupling can impose later: abandonment, a pricing pivot, a license change, security posture, and exit cost. A dependency adopted without that weighing is an unpriced liability.

• Building what already exists is a finding. Re-implementing a solved,

well-served problem (native or vetted) is a selection error in the other direction — name it.

Mandatory routing — no verdict from memory

When the selection is load-bearing, the field survey and the maintenance / security / license / adoption judgement must come from CURRENT research, not training-data recall — a tool's maintenance status, licensing, and security posture drift constantly. Route to a research capability rather than judging from memory: /discovery:research, or /discovery:research-deep for a large surface. Degrade to an explicit in-thread research pass (fetch the primary sources yourself and cite them) when that capability is not installed — never a bare recalled verdict.

Audit — what to look for

Name concrete, located findings (per the method doc's step 2). This fires both at an explicit choice-time AND over choices already embedded in the work in flight:

• a tool named before the problem it solves was stated;
• a single candidate with no field surveyed;
• a choice tracing to habit, availability, incumbency, or preconception

rather than fit;

• a third-party dependency adopted with no maintenance / security / license /

exit-cost weighing;

• a native or authoritative option skipped straight to third-party;
• a hand-built solution to an already-solved, well-served problem.

Correct each forward now: re-derive the choice from the stated problem, run (or route) the survey, walk the ladder, and price the coupling. Where the evaluation is load-bearing, route it to research rather than self-checking from the same recall that produced the reflex.

Distinct axis — the incumbency overlap

The incumbency sin here is selection-specific — "the repo uses X, so this new work uses X". The general interrogation of inherited precedent is /discipline:reason-dont-recite; this skill applies that lens narrowly to a technology choice and adds the field survey, the ladder, and the coupling price. Route a broader inherited-design challenge there.

What this skill does NOT do

• Does not churn a well-fitted choice. A tool re-derived from the problem

and found to fit audits clean — including an incumbent one; the duty is to re-derive, not to switch for its own sake.

• Does not issue a verdict from recall. A load-bearing evaluation routes

to research; the skill does not pronounce a maintenance or security verdict from memory.

• Does not fabricate a finding. A choice already problem-derived and

field-surveyed audits clean; say so.

Gotchas

• The reflex hides as a reason. "We use X here" and "X is what I know"

are the incumbency and habit sins wearing the costume of a rationale; a real rationale names the requirement X satisfies, not its familiarity.

• A deep dependency-inventory variant is deliberately deferred. Auditing

an entire repo's dependency graph for coupling risk is out of scope here; that sibling lands on the first real dependency-inventory audit request, not before — not shipped speculatively.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: melodic-software
• Source: melodic-software/claude-code-plugins
• License: MIT

Install and usage instructions live in the source repository linked above.