learn

# /learn — Knowledge Compiler

## Orientation

**Use when:** You have a completed campaign or evolve cycle and want to compile
its findings into the project's growing knowledge wiki — so future sessions
start smarter, not from scratch.

**Don't use when:** You want a structured incident analysis first (use `/postmortem`
— run it before `/learn`); you haven't finished any campaigns (nothing to compile);
you want a context transfer only (use `/session-handoff`).

**Key difference from appending:** `/learn` doesn't create isolated per-campaign
files. It integrates new findings into existing wiki pages — updating evidence
lists, raising confidence where a pattern is confirmed again, and flagging
contradictions. A wiki is a compiler; a log is an interpreter.

## Invocation Forms

```
/learn                              — most recently completed campaign
/learn {slug}                       — specific campaign by slug
/learn {file-path}                  — specific campaign file path
/learn --from-evolve {target}       — compile from /evolve pattern library
/learn --from-evolve {target} --cycle {n}  — specific evolve cycle only
/learn --lint                       — lint-only pass (no new extraction)
/learn --compile                    — re-compile staging area into wiki (no new extraction)
/learn --memory                     — compile semantic memory blocks from planning artifacts
/learn --doc-sync                   — process doc-sync queue into .planning/doc-sync/latest.md
```

## Inputs

1. A campaign slug, file path, evolve target, or "most recent" resolution
2. Corresponding postmortem in `.planning/postmortems/` (optional)
3. `.planning/telemetry/audit.jsonl` filtered to this campaign (optional)
4. For `--from-evolve`: `.planning/evolve/{target}/pattern-library.md`

## Protocol

### Step 1: RESOLVE TARGET

**If `/learn` (no argument):**
- Glob `.planning/campaigns/completed/*.md` or `.planning/campaigns/*.md`
  where `Status: completed`
- Sort by modification time descending, take most recent
- If none found: "No completed campaigns found. Run /learn after a campaign completes." Stop.

**If `/learn {slug}`:**
- Search `.planning/campaigns/` for a file whose name contains `{slug}`
- Check `.planning/campaigns/completed/` if not found in active
- If still not found: "No campaign found matching '{slug}'."

**If `/learn --from-evolve {target}`:**
- Read `.planning/evolve/{target}/pattern-library.md` — this is the source
- If `--cycle {n}`: filter to sections beginning with `## Cycle {n}` only
- If file not found: "No evolve pattern library for '{target}'." Stop.

**If `/learn --doc-sync`:** Run:
```
node hooks_src/doc-sync.js
```
Then review `.planning/doc-sync/latest.md`. Stop after reporting the queue count,
files surfaced, skipped deleted files, and report path. Do not run campaign
extraction unless the user separately asks for it.

**If `/learn --lint`, `/learn --compile`, or `/learn --memory`:** Skip to Step 4, 5, or 5.5 respectively.

### Step 2: GATHER SOURCES

For campaign-based runs only (skip for `--from-evolve`):

**Campaign file (required):**
- Full content — direction, phases, Decision Log, circuit breaker activations

**Postmortem (optional):**
- Search `.planning/postmortems/` for files matching `*{slug}*`
- If not found: note "Postmortem not found — proceeding without it" and continue

**Audit telemetry (optional):**
- Read last 200 lines of `.planning/telemetry/audit.jsonl`
- Filter entries that match the campaign slug or its active period
- If none: note "No audit telemetry found for this campaign"

### Step 3: FLUSH

Extract raw findings and write to staging.

**For campaign sources:**

Extract four categories:

**A. Successful Patterns** — approaches that demonstrably worked (phases completed without rework, postmortem positives, no reverts).
Per pattern: `name`, `mechanism` (what caused success), `evidence` (phase/commit/entry), `topic` (infer from subject matter), `applicability`.

**B. Anti-patterns** — what was tried and failed (rework phases, circuit breaker trips, quality gate blocks, reverts).
Per pattern: `name`, `what-was-tried`, `failure-mode`, `evidence`, `topic`, `avoidance`.

**C. Key Decisions** — from Decision Log or inferred from phase descriptions.
Per decision: `what`, `rationale`, `outcome` (completed or rework).

**D. Quality Rule Candidates** — only generate if: specific regex, applies to a specific file pattern, occurred more than once or was severe. Per candidate: regex, file pattern, trigger message, confidence (`high`/`medium`/`low` — skip `low`).

**For evolve sources:**

Parse the pattern library's sections. For each pattern record:
- `name`: section heading
- `mechanism`: "**Mechanism:**" field
- `delta`: "**Delta:**" field
- `topic`: infer from "**Axis class:**" (e.g., `orientation_precision` → `skill-orientation`)
- `applies-to`: "**Applies to:**" field
- `confidence`: "**Confidence:**" field
- `evidence`: source file + cycle number

**Staging write:**

Create `.planning/wiki/_staging/` if it does not exist.
Write staged findings to `.planning/wiki/_staging/{source-slug}-{timestamp}.jsonl` —
one JSON record per finding (newline-delimited).

If zero findings are extractable: write staging file with a single `{"type":"empty","source":"{slug}"}` record and note "Campaign may have been too brief."

### Step 4: COMPILE

Integrate staged findings into wiki pages.

Create `.planning/wiki/` if it does not exist.

For each staged finding:
1. Determine the wiki page: `.planning/wiki/{topic}.md` where `topic` is the finding's `topic` field (normalized to kebab-case).
2. Read the wiki page if it exists.
3. **If the page exists and contains a section for this pattern** (match on `## {name}`):
   - Append the new source to the `**Evidence:**` list
   - Update `**Last confirmed:**` to today
   - If new confidence >= existing confidence: raise it
   - If new evidence contradicts the existing mechanism: add a `**Conflict:**` field — do not silently overwrite
4.