@dzhechkov/harness-cli 0.2.2 → 0.2.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +36 -8
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -130,39 +130,67 @@ npx @dzhechkov/keysarium init
|
|
|
130
130
|
**Goal:** Continuously gather data, add new sources, and "recalculate" the product vision as insights accumulate.
|
|
131
131
|
|
|
132
132
|
```bash
|
|
133
|
-
# Install keysarium +
|
|
133
|
+
# Install keysarium (research pipeline) + evidence-wiki (knowledge base)
|
|
134
134
|
npx @dzhechkov/keysarium init
|
|
135
|
+
# Copy evidence-wiki plugin into your project:
|
|
136
|
+
npx @dzhechkov/evidence-wiki # or git clone https://github.com/djd1m/evidence-wiki
|
|
137
|
+
|
|
135
138
|
npm install -g @dzhechkov/harness-cli
|
|
136
139
|
dz init --target claude-code --preset meta
|
|
137
|
-
|
|
138
|
-
# Additionally install the memory package for persistent learning:
|
|
139
|
-
# (already included if using keysarium)
|
|
140
140
|
```
|
|
141
141
|
|
|
142
|
-
**Workflow — iterative research cycles:**
|
|
142
|
+
**Workflow — iterative research cycles with evidence wiki:**
|
|
143
143
|
|
|
144
144
|
```
|
|
145
145
|
Week 1: /casarium "Product X — initial research"
|
|
146
146
|
→ researches/ directory created with findings
|
|
147
147
|
→ .keysarium/memory/ stores patterns + reward scores
|
|
148
148
|
|
|
149
|
+
/wiki-generate ← evidence-wiki
|
|
150
|
+
→ Scans researches/, ADRs, docs
|
|
151
|
+
→ Generates wiki/concepts/*.md (atomic pages with inline sources)
|
|
152
|
+
→ Builds wiki/graph.json (knowledge graph)
|
|
153
|
+
→ wiki/INDEX.md links everything
|
|
154
|
+
|
|
149
155
|
Week 2: Add new data → /casarium "Product X — update with Q2 metrics"
|
|
150
156
|
→ Memory recalls Week 1 patterns (reward-calibrated learning)
|
|
151
157
|
→ New findings merged with existing, conflicts resolved
|
|
152
|
-
|
|
158
|
+
|
|
159
|
+
/wiki-generate --check ← re-generates wiki
|
|
160
|
+
→ New concepts added, existing updated
|
|
161
|
+
→ Every claim verified: triple-pillar protocol requires N independent
|
|
162
|
+
typed sources (ADR + methodology + research)
|
|
163
|
+
→ Stale concepts flagged, broken evidence links detected
|
|
164
|
+
|
|
165
|
+
/triple-check wiki/concepts/pricing-model.md ← verify specific page
|
|
166
|
+
→ Checks that every factual claim has inline source citations
|
|
167
|
+
→ Flags unsupported statements
|
|
153
168
|
|
|
154
169
|
Week N: /casarium "Product X — pivot analysis after customer feedback"
|
|
155
|
-
→ Full history
|
|
170
|
+
→ Full history in memory layer + evidence wiki
|
|
156
171
|
→ /harvest extracts reusable knowledge patterns
|
|
172
|
+
→ /wiki-generate rebuilds the entire knowledge graph
|
|
173
|
+
→ Product vision "recalculated" — the wiki IS the living product model
|
|
157
174
|
```
|
|
158
175
|
|
|
176
|
+
**The evidence-wiki advantage:**
|
|
177
|
+
|
|
178
|
+
| Without evidence-wiki | With evidence-wiki |
|
|
179
|
+
|----------------------|-------------------|
|
|
180
|
+
| Research in markdown files | Atomic concept pages with inline sources |
|
|
181
|
+
| Findings scattered across `researches/` | Interlinked knowledge graph (`graph.json`) |
|
|
182
|
+
| "I think we decided X" | Every claim has a cited source (triple-pillar) |
|
|
183
|
+
| Hard to see what changed | `/wiki-generate --check` diffs the knowledge base |
|
|
184
|
+
| No verification | `/triple-check` enforces evidence discipline |
|
|
185
|
+
|
|
159
186
|
**Key features for long-term research:**
|
|
187
|
+
- **Evidence wiki** (`@dzhechkov/evidence-wiki`): atomic concept pages where every factual claim carries inline sources; knowledge graph for cross-referencing; triple-pillar protocol (N independent typed sources per claim)
|
|
160
188
|
- **Reward-calibrated memory** (`@dzhechkov/memory` Reflexion): each checkpoint response trains the system — "ок" = excellent (1.0), feedback = good (0.7), rework = needs_work (0.3)
|
|
161
189
|
- **Agent SDK Dreaming**: between sessions, patterns are consolidated and distilled
|
|
162
190
|
- **`/harvest`** (knowledge-extractor skill): extracts reusable patterns from completed research into `lib/` templates
|
|
163
191
|
- **SQLite + FTS5 backend**: scales to 100k+ records with full-text search across all research sessions
|
|
164
192
|
|
|
165
|
-
**Best for:** Product strategy over months, continuous market monitoring, evolving product vision.
|
|
193
|
+
**Best for:** Product strategy over months, continuous market monitoring, evolving product vision with evidence-backed decisions.
|
|
166
194
|
|
|
167
195
|
---
|
|
168
196
|
|
package/package.json
CHANGED