Fleet Canon Burn-Down Plan — COMPLETE
Status: 43/43 agents certified conformant (2026-06-09)
Final scorecard
| Cycle | PR | Agents upskilled | Running total |
|---|---|---|---|
| Baseline | — | engineering-director, agent-auditor | 2 |
| Cycle 1 (batch-1 + batch-2) | #123 | COO, CGEO, analytics-engineer, CPTO, data-analytics-director, design-research-director, documentarian | 9 |
| Cycle 2 (T1) | #124–126 | staff-engineer, platform-engineer, security-analyst | 12 |
| Cycle 3 (T1) | #129 | ai-engineer, ai-systems-director, human-resources, innovation-director, process-engineer, project-manager, security-director, site-reliability-engineer | 20 |
| T2 batch-1 | #131 | brand-strategist, community-manager, competitive-intelligence-analyst, compliance-analyst, content-strategist, creative-director, customer-success-director, design-research-director, external-reviewer, finance-director, financial-analyst | 31 |
| T2 batch-2 | #130 | golden-rule, growth-director, historian, instructional-designer, legal-director, product-manager, quality-risk-director, ux-designer, ux-research-associate, video-producer, visual-designer | 42 |
| MCP agents | #132 | archivist, personal-assistant | 43/43 |
What “certified conformant” means
- T1: charter + versioned prompt + TOOLSPEC with Bash denylist + MCP tool inventory (where applicable) + two-tier eval (contract gate CI + behavioural cases) + currency (model.lock, CHANGELOG, RUNBOOK)
- T2: charter + versioned prompt + TOOLSPEC (Edit/Write blast radius, no Bash) + eval + currency
Remaining work
- Behavioural eval baseline runs for all agents (Tier-2 on-demand; not CI-gated)
- Future: when a new agent is hired, run canon-upskill-cycle workflow from day one