| name | author-gap-scan |
| description | Audit an existing or proposed track DAG for foundational gaps — concepts referenced but never taught, orphan roots that assume prior knowledge, prerequisite leaps, redundant nodes, and dangling supplementary references. Produces a prioritized gap report and an authoring plan to close them. Trigger after author-structure, before a milestone hand-off, or when the user says "gap scan", "coverage check", "what's missing", "any foundational gaps", or "/author-gap-scan". Spawns coverage-auditor. Read-only analysis — proposes fixes, edits nothing. |
author-gap-scan — foundational-gap audit
You find the gaps that make self-teaching fail — the exact failure the product exists to prevent
(README.md: a learner uses something for months yet can't explain a foundational concept it rests
on). You analyze and report; you do not edit the graph or draft nodes.
Procedure
- Spawn
coverage-auditor on the track directory (or a concept-mapper proposal). It
returns findings keyed to node ids/handles: untaught concepts, orphan roots, prerequisite
leaps, redundancy, dangling supplementary refs, and motivation gaps.
- Prioritize. Rank by learner impact — untaught-concept and prereq-leap findings are
top priority (they are the silent gaps); redundancy and dead-weight glossary terms are lower.
- Turn findings into an authoring plan, not just a list:
- untaught concept → "add a root/intermediate node teaching X; wire it as a prerequisite of Y".
- orphan root assuming prior knowledge → "either add the missing prerequisite node, or confirm
this is a legitimate entry point."
- prereq leap → "insert an intermediate node between src and dst."
- redundancy → route to
author-restructure (merge is a restructure, not an ad-hoc delete —
the directory only grows; pieces are tombstoned, never removed).
- motivation gap → route to
author-snippet to add why it matters / what it unlocks.
- Draw the boundary with the CLI. The deterministic
synaptic validate catches structural
breakage (unresolved ids/slugs, cycles). This skill catches the pedagogical gaps validate
cannot see — a graph can be perfectly valid and still teach a concept it never grounded. Say
which findings the CLI would also catch and which are yours alone.
Output
A prioritized gap report + a concrete authoring plan (which nodes to add and where to wire them),
handed to the human. Loop back to author-structure to add the missing nodes, then re-scan.
Hard rules
- Read-only: propose fixes; never edit the DAG, mint ids, or delete anything.
- Never push or commit.