@fredericboyer/dev-team 0.8.1 → 0.10.0

package/templates/CLAUDE.md

@@ -19,9 +19,9 @@ This project uses [dev-team](https://github.com/dev-team) — adversarial AI age
 | `@dev-team-brooks` | Architect & Quality Reviewer | Architectural review, coupling, ADR compliance, quality attributes (performance, maintainability, scalability) |
 | `@dev-team-conway` | Release Manager | Versioning, changelog, release readiness, semver validation |
 | `@dev-team-drucker` | Team Lead / Orchestrator | Auto-delegates to specialists, manages review loops, resolves conflicts |
-| `@dev-team-borges` | Librarian | End-of-task/review/audit memory review, cross-agent coherence, system improvement |
+| `@dev-team-borges` | Librarian | End-of-task memory extraction, cross-agent coherence, system improvement |
 
-### Workflow
+### Capabilities
 
 For automatic delegation, use `@dev-team-drucker` — it analyzes the task and routes to the right specialist.
 
@@ -34,15 +34,17 @@ For non-trivial work: explore the area first, then implement, then review.
 - **Hamilton** — auto-flagged when infrastructure/operations files change (Dockerfile, docker-compose, CI workflows, Terraform, Helm, k8s, health checks, monitoring config, .env templates, etc.)
 - **Voss** — auto-flagged when app config/data files change (.env, config, migrations, database, etc.)
 - **Deming** — auto-flagged when tooling files change (eslint, CI workflows, package.json, etc.)
-- **Tufte** — auto-flagged when documentation files change (.md, /docs/, README, etc.) AND when significant implementation files change (src/, templates/agents/, templates/skills/, templates/hooks/, bin/, package.json) to detect doc-code drift
+- **Tufte** — auto-flagged when documentation files change (.md, /docs/, README, etc.) AND when significant implementation files change to detect doc-code drift
 - **Brooks** — auto-flagged when any non-test implementation code changes (quality attributes) and when architectural boundaries are touched (/adr/, /core/, /domain/, /lib/, build config, etc.)
 - **Conway** — auto-flagged when release artifacts change (package.json, changelog, version files, release/publish/deploy workflows, etc.)
 
 **End-of-workflow agents:**
-- **Borges** — mandatory at end of every `/dev-team:task`, `/dev-team:review`, `/dev-team:audit`, and `/dev-team:assess`. Reviews memory freshness, cross-agent coherence, and system improvement opportunities.
+- **Borges** — mandatory at end of every `/dev-team:task`, `/dev-team:review`, `/dev-team:audit`, and `/dev-team:assess`. Extracts structured memory entries, reviews cross-agent coherence, and identifies system improvement opportunities.
 
 **Orchestration:**
-- **Drucker** — delegates tasks to the right implementing agent and spawns reviewers. Szabo, Knuth, and Brooks review all code changes. Brooks covers both structural review and quality attribute assessment (performance, maintainability, scalability).
+- **Drucker** — delegates tasks to the right implementing agent and spawns reviewers. Szabo, Knuth, and Brooks review all code changes.
+
+**CRITICAL: Always run agents in the background.** When spawning Drucker or any agent for tasks that take more than a few seconds, use `run_in_background: true`. The main conversation loop must remain interactive.
 
 Agents challenge each other using classified findings:
 - `[DEFECT]` blocks progress. `[RISK]`, `[QUESTION]`, `[SUGGESTION]` are advisory.
@@ -50,7 +52,7 @@ Agents challenge each other using classified findings:
 
 ### Parallel execution
 
-When working on multiple independent issues, use parallel agents on separate branches. Drucker coordinates the review wave after all implementations complete. See ADR-019 for the full model: Brooks assesses file independence, implementations run concurrently, reviews are batched into a coordinated wave, defects route back per-branch, and Borges runs once across all branches at the end.
+When working on multiple independent issues, use parallel agents on separate branches. Drucker coordinates the review wave after all implementations complete.
 
 ### Hook directives are MANDATORY
 
@@ -63,25 +65,37 @@ Do NOT skip this. Do NOT treat hook output as optional. If you believe a review
 
 ### Skills
 
+**Framework skills** (installed automatically, updated with `dev-team update`):
 - `/dev-team:challenge` — critically examine a proposal or implementation
 - `/dev-team:task` — start an iterative task loop with adversarial review gates
 - `/dev-team:review` — orchestrated multi-agent parallel review of changes
 - `/dev-team:audit` — full codebase security + quality + tooling audit
-- `/dev-team:merge` — merge a PR with Copilot review handling, auto-merge, CI monitoring, and post-merge actions
-- `/dev-team:security-status` — check code scanning, Dependabot, and secret scanning alerts
 - `/dev-team:assess` — audit knowledge base health (learnings, agent memory, CLAUDE.md)
 
-### Learnings — where to write what
+**Optional workflow skills** (installed to `.claude/skills/` during init, not overwritten on update):
+- Check `.claude/skills/` for project-specific workflow skills (merge automation, security monitoring, etc.)
+
+### Memory architecture (two-tier)
 
 All project and process learnings MUST go to in-repo files, NOT to machine-local memory (`~/.claude/projects/`). Machine-local memory is invisible to other developers, agents, and sessions.
 
+**Tier 1 — Shared team memory** (`.dev-team/learnings.md`):
+Project facts, overruled challenges, cross-agent decisions, process rules. All agents read this at session start.
+
+**Tier 2 — Agent calibration memory** (`.dev-team/agent-memory/<agent>/MEMORY.md`):
+Domain-specific findings, known patterns, active watch lists. Each agent owns its own file. Entries include `Last-verified` dates for temporal decay.
+
 | What | Where | Examples |
 |------|-------|---------|
-| Project patterns, process rules, tech debt, overruled challenges | `.dev-team/learnings.md` | "We use PostgreSQL", "Hooks over guidelines", "Knuth's finding X was overruled because Y" |
-| Agent-specific calibration | `.dev-team/agent-memory/<agent>/MEMORY.md` | Szabo: "Auth uses JWT not sessions", Knuth: "Coverage weak in parsers" |
+| Project patterns, process rules, tech debt, overruled challenges | `.dev-team/learnings.md` (Tier 1) | "We use PostgreSQL", "Hooks over guidelines", "Knuth's finding X was overruled because Y" |
+| Agent-specific calibration | `.dev-team/agent-memory/<agent>/MEMORY.md` (Tier 2) | Szabo: "Auth uses JWT not sessions", Knuth: "Coverage weak in parsers" |
 | Formal architecture decisions | `docs/adr/` | ADR format, not learnings |
 | User-specific preferences only | Machine-local memory | Personal style, name, role — things that vary per person, not per project |
 
+**Memory evolution:** New entries trigger re-evaluation of related existing entries. Duplicates are merged, contradictions are superseded, and 3+ overrules on the same tag generate calibration rules.
+
+**Temporal decay:** Entries have `Last-verified` dates. Borges flags entries not verified in 30+ days and archives entries over 90 days to the `## Archive` section.
+
 When the human gives feedback about process, coding style, or tool behavior: write it to `.dev-team/learnings.md`. Only use machine-local memory for things that are truly personal and would not apply to another developer on the same project.
 
 <!-- dev-team:end -->

package/templates/agent-memory/dev-team-beck/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Beck (Test Implementer)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Test Patterns and Conventions
-
-
-## Framework and Runner Notes
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-borges/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Borges (Librarian)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Memory Health Status
-
-
-## System Improvement Log
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Recommendations accepted/deferred — tunes what to flag over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-brooks/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Brooks (Architect)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Project Conventions
-
-
-## Patterns to Watch For
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-conway/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Conway (Release Manager)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Project Conventions
-
-
-## Patterns to Watch For
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-deming/MEMORY.md

@@ -1,12 +1,26 @@
-# Agent Memory: Deming (Tooling & DX Optimizer)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+# Agent Memory: Deming (Tooling Optimizer)
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Tooling Decisions
-
-
-## Hook Effectiveness
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-drucker/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Drucker (Orchestrator)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Delegation Patterns
-
-
-## Conflict Resolution Log
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Delegation decisions that worked well or poorly — tunes routing over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-hamilton/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Hamilton (Infrastructure Engineer)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Project Conventions
-
-
-## Patterns to Watch For
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-knuth/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Knuth (Quality Auditor)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Coverage Gaps Identified
-
-
-## Recurring Boundary Conditions
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-mori/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Mori (Frontend/UI Engineer)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Project Conventions
-
-
-## Patterns to Watch For
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-szabo/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Szabo (Security Auditor)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Trust Boundaries Mapped
-
-
-## Known Attack Surfaces
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-tufte/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Tufte (Documentation Engineer)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Project Conventions
-
-
-## Patterns to Watch For
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agent-memory/dev-team-voss/MEMORY.md

@@ -1,12 +1,26 @@
 # Agent Memory: Voss (Backend Engineer)
-<!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Tier 2: Agent calibration memory. Domain-specific findings, patterns, and watch lists. -->
+<\!-- First 200 lines are loaded into agent context. Keep concise. -->
+<\!-- Borges extracts structured entries automatically after each task. -->
 
-## Project Conventions
-
-
-## Patterns to Watch For
+## Structured Entries
+<\!-- Format:
+### [YYYY-MM-DD] Finding summary
+- **Type**: DEFECT | RISK | SUGGESTION | OVERRULED | PATTERN | DECISION
+- **Source**: PR #NNN or task description
+- **Tags**: comma-separated relevant tags
+- **Outcome**: accepted | overruled | deferred | fixed
+- **Last-verified**: YYYY-MM-DD
+- **Context**: One-sentence explanation
+-->
 
+## Calibration Rules
+<\!-- Auto-generated when 3+ findings on the same tag are overruled. -->
+<\!-- Format: "Reduce severity for [tag] findings — overruled N times (reason)" -->
 
 ## Calibration Log
-<!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
+<\!-- Challenges accepted/overruled — tunes adversarial intensity over time -->
 
+## Archive
+<\!-- Entries older than 90 days without verification are moved here by Borges. -->
+<\!-- Not loaded into agent context but preserved for reference. -->

package/templates/agents/dev-team-beck.md

@@ -14,6 +14,8 @@ Your philosophy: "Red, green, refactor — in that order, every time."
 
 **Memory hygiene**: Read your MEMORY.md at session start. Remove stale entries (overruled challenges, outdated patterns). If approaching 200 lines, compress older entries into summaries.
 
+**Role-aware loading**: Also read `.dev-team/learnings.md` (Tier 1). For cross-agent context, scan entries tagged `testing`, `coverage`, `boundary-condition`, `test-pattern` in other agents' memories — especially Knuth (quality findings to implement) and Voss/Mori (implementation patterns to test).
+
 Before writing tests:
 1. Spawn Explore subagents in parallel to understand existing test patterns, frameworks, and conventions in the project.
 2. **Research current practices** when choosing test frameworks, assertion libraries, or testing patterns. Check current documentation for the test runner and libraries in use — APIs change between versions, new matchers get added, and best practices evolve. Prefer codebase consistency over newer approaches; flag newer alternatives as `[SUGGESTION]` when they do not fit the existing conventions.
@@ -58,6 +60,7 @@ Rules:
 3. When challenged: address directly, concede when wrong, justify with a counter-scenario when you disagree.
 4. One exchange each before escalating to the human.
 5. Acknowledge good work when you see it.
+6. **Silence is golden**: If you find nothing substantive to report, say "No substantive findings" and stop generating additional findings. You must still complete the mandatory MEMORY.md write and Learnings Output steps. Do NOT manufacture `[SUGGESTION]`-level findings to fill the review. A clean review is a positive signal, not a gap to fill.
 
 ## Learning