knowledge

knowledge contains the three skills that keep a living knowledge base useful over time. It is for teams that have an established knowledge/ tree and want dedicated operations over it — without necessarily adopting the full audit surface of quality.

Why this exists

A knowledge base degrades quietly. Entries go stale as the codebase evolves. Cross-domain connections that would help the next implementer never get recorded. Contradictions accumulate between modules that were written months apart. And when a new teammate (or agent) starts, there is no signal about which knowledge is fresh and which is archaeology.

These three skills target that degradation at the maintenance, synthesis, and briefing layers respectively. They are not audit gates — they are operational routines that keep the knowledge loop healthy.

What it includes

skills:
  - knowledge-hygiene    # staleness, orphans, broken links, archive drift
  - knowledge-synthesize # cross-node connections, forgotten context, contradictions
  - knowledge-briefing   # what changed, what needs attention, proactive digest
roles: []

Source: bundles/knowledge.yml

Why each one is in

• knowledge-hygiene — scans the knowledge tree for stale entries (facts that contradict current code), orphaned nodes (no incoming or outgoing links), broken references, and archive drift (entries promoted to canonical status that are no longer accurate). Produces a prioritised list of items that need review or retirement.
• knowledge-synthesize — reads across the knowledge tree to find connections that were never recorded: a fact in knowledge/payments that clarifies a pattern in knowledge/auth, a forgotten-but-relevant entry that the current implementer would want to know, contradictions between two nodes that have diverged. Surfaces these as proposed additions or corrections.
• knowledge-briefing — generates a proactive digest of what has changed in the knowledge base since a given point (last session, last week, last release) and flags entries that need a human decision. Designed to be run at the start of a session or a sprint to orient the agent and the team.

Why not included

• audit-grounding — also reads the knowledge tree, but it does so as a signal-only divergence check tied to the code diff. It belongs in quality because its trigger and scope are diff-centric, not knowledge-tree-centric.
• doc-lifecycle, doc-adr — those produce knowledge; these three consume and maintain it. Producers live in starter and docs.

Workflow this enables

These skills map to three distinct rhythms:

• On-demand maintenance — run /octopus:knowledge-hygiene after a large feature lands to catch entries that the new code has made stale. Takes a path (defaults to knowledge/) and produces a prioritised fix list.
• Periodic synthesis — run /octopus:knowledge-synthesize quarterly or after merging a cross-domain feature. The skill proposes new connections and surfaces contradictions; you decide which ones to record.
• Session or sprint start — run /octopus:knowledge-briefing to orient yourself (or a fresh agent session) before picking up a task. The skill summarises what has changed and flags any entries that need a decision before work begins.

Composes with

• starter — provides doc-adr and doc-lifecycle, which produce the knowledge entries this bundle maintains.
• docs — for the full documentation lifecycle (RFC → Spec → ADR → Knowledge → Changelog) that feeds the knowledge tree.
• quality — includes the audit gates and advisory signals alongside which knowledge provides the knowledge-loop operations.