speclisted

# spec — analyze and execute a spec Load at the start of any session that builds or implements a feature, whether or not the work is framed as a "spec". A spec governs the work -> this skill executes it; one should exist but doesn't -> the `docs` skill authors it first. Run **Analyze** before touching any code. Blockers / unclear items you can't resolve alone -> pause for the FnB. `<self>` = your shell_id. --- ## Step 1: Load the spec A feature can hold several unfrozen specs at once (see the `docs` skill). NEVER auto-pick "the latest" — list the feature's open specs and choose the target explicitly: ``` # the feature's documents — pick an unfrozen spec (frozen=0) by id: sc mem get documents --feature <id> # load the chosen spec body: sc mem get documents --doc <doc_id> # the spec's task plan (empty = no plan yet): sc mem get tasks --doc <doc_id> ``` `get documents --feature <id>` lists every spec/doc with `kind`, `seq`, `frozen`, `task_count`. Active spec = the unfrozen one with `task_count > 0` — resume it. `task_count = 0` = backlog; starting it (Step 3) makes it active. More than one open spec and the target unclear -> ask the FnB. Tasks already exist -> skip to **Step 4** (Track). Read the entire spec body before going further. Do not skim. --- ## Step 2: Analyze Surface all three before any planning or code: ### Viability - Session-completable? Bounded + clear entry points = yes. Multiple layers / migrations / unknown dependencies = no -> say so + propose