spek-knowledge

name	spek-knowledge
description	Search, contribute to, or update the project's knowledge base.

What this skill does

This skill orchestrates the existing {{command}} knowledge CRUD surface for ad-hoc read, contribute, and update operations on the project's knowledge store, without starting a spec/plan/implement flow. Unlike spek-new, spek-plan, and spek-implement, it does not drive an interactive CLI state machine — it is a static playbook. The agent recognises the user's natural-language intent, picks one of three branches (lookup / contribute / update), and calls the matching {{command}} knowledge command directly.

When to invoke

Invoke this skill any time the user references the knowledge base, an entry, a convention worth remembering, or asks a question that the knowledge store might already answer. Typical natural-language triggers include:

• "What do we know about X?"
• "Search the knowledge base for Y."
• "Remember that Z is the case." / "Add a note about Z."
• "Update what we have on W."
• "Recall the convention for V."

One skill handles all three intents. Discriminate by what the user actually said — do not ask the user to pick a slash command per intent.

Intent: lookup

Triggered when the user wants to read or search existing entries. A lookup does not dump the raw hit list — it returns a single consolidated, source-cited answer with duplicates removed and the most specific source winning. The flow has a deterministic stage (exact de-dup) and a judgement stage (consolidation), kept strictly separate.

1. Search. Run {{command}} knowledge search <query> with a concise query derived from the user's question. The output is a ranked list of results — one per matching document, strongest match first — each carrying its scope (e.g. project, team, global), path, title, score, category (the kind of knowledge: e.g. gotchas, architecture, learnings), checksum (a content hash), and up to three excerpts. A document matches when every query word occurs somewhere in it, in any order. If there are no hits, say so plainly and stop — do not fall back to a write unless the user explicitly asks to add a new entry.

2. Exact de-dup (deterministic — no judgement). Group the hits by their checksum. Hits sharing a checksum are byte-identical copies of the same entry held in more than one place; collapse each such group to a single candidate, unioning the scope/path citations of every copy in the group. This is pure equality — never merge two entries whose checksums differ at this stage, however similar they look. The result is a list of unique candidates, each with one or more source citations.

3. Consolidate (judgement — delegated to a sub-agent). Hand the unique candidates to a consolidation sub-agent so the raw bodies never crowd the main context. The sub-agent's contract:

   • Input: the user's question and the list of unique candidates (each with its scope(s), path(s), and category).
   • Task: read each candidate's full body with {{command}} knowledge read --data '{"scope":"<scope>","path":"<path>"}', then classify the relationship between candidates and combine them:
     • Equivalent (same point, different words) → merge into one point, citing every source.
     • Refinement (one is a more specific case of another) → apply layered precedence: the most-specific scope wins. Specificity runs project (most specific) → team → global (least specific); a project entry overrides a team entry overrides a global one for the same item. State the winning guidance and note what it overrode.
     • Genuine contradiction (sources actually disagree) → surface it explicitly as a conflict naming both sources; never silently drop or average it.
     • Distinct (unrelated points) → keep both.
   • Output (returned to the main agent): a single consolidated answer composed of merged points, each citing the source scope(s)/path(s) it was drawn from, with any contradictions presented as surfaced conflicts. The raw per-source candidate list is not the output.

4. Present the sub-agent's consolidated answer to the user, keeping every citation (scope + path) visible so the user can see which configured store each point came from. Never present the raw hit list as the result.

If the executing agent cannot spawn a sub-agent, run the exact same consolidation inline in the main context instead: read the unique candidates' bodies, apply the identical relationship-classification and layered-precedence rules, and present the same single cited answer. The output is identical; only the context isolation is weaker. Do not block on the absence of sub-agent orchestration.

Intent: contribute

Triggered when the user wants to record something new.

1. Run {{command}} knowledge sources to enumerate configured scopes (the authoritative list of writable destinations), and {{command}} knowledge categories to load the category definitions — each category's purpose, boundary, retrieval tier, and expected entry shape.
2. Route the entry to a category. From the definitions, pick the category whose Purpose matches the entry and whose Boundary does not push it elsewhere — e.g. a standing rule is a convention, a defined term is a glossary entry, the reasoning behind a choice is a decision, an empirical finding is a learning, a structural fact is architecture, a sharp edge is a gotcha. Honour the entry shape: the glossary is for a term and a short gloss only — steer over-long or multi-paragraph content to a more fitting category (architecture, learnings, decisions) rather than letting it bloat the always-applied glossary. The entry's path is then <category>/<slug>.md, a slug-style filename under the chosen category.
3. Decide (or ask the user) which scope to write to and the entry body. The path comes from the category routing in step 2.
4. Stage the body on disk under .spektacular/tmp/<slug>.md using the Write tool. Do not pipe the body via stdin; the only supported invocation is --file <staged>.
5. Show the user the proposed scope, path (category + slug), and body before writing. Wait for explicit confirmation ("yes", "go ahead", or equivalent). If the user asks for changes, revise the staged body and re-show — never write on an implicit signal.
6. Only after explicit confirmation, run:

   {{command}} knowledge write --data '{"scope":"<scope>","path":"<category>/<slug>.md"}' --file .spektacular/tmp/<slug>.md
   

7. Remove the scratch file after a successful write: rm .spektacular/tmp/<slug>.md.

Intent: update

Triggered when the user wants to revise an existing entry.

1. Identify the target entry. Run {{command}} knowledge search <query> (or read the user-supplied path directly) to locate it. Confirm the scope and path with the user if there is any ambiguity.
2. Read the current body with {{command}} knowledge read --data '{"scope":"<scope>","path":"<path>"}'.
3. Apply the user's revision intent to produce new content. Stage the revised body under .spektacular/tmp/<slug>.md using the Write tool.
4. Show the user the scope, path, and proposed new body (or a diff against the current body) before writing. Wait for explicit confirmation.
5. Only after explicit confirmation, run:

   {{command}} knowledge write --data '{"scope":"<scope>","path":"<path>"}' --file .spektacular/tmp/<slug>.md
   

   The scope and path must match the original — that is what makes this an update rather than a new entry.
6. Remove the scratch file after a successful write: rm .spektacular/tmp/<slug>.md.

Decline handling

If the user declines, asks for changes, or expresses uncertainty at any propose-then-confirm checkpoint, do not invoke {{command}} knowledge write. Either loop back to refine the proposal — adjust scope, path, or body and re-show — or stop and leave the knowledge store untouched. Removing the staged scratch file at .spektacular/tmp/<slug>.md is fine either way; a half-finished proposal should not linger on disk.

The propose-then-confirm contract is enforced by this prose, not by a CLI guard. Treat it as load-bearing: a write without explicit user approval is a bug in the skill's execution, not an acceptable shortcut.