Agentic Development · Live coverage

Did the governance loop close? Here is the daily, reproducible measurement.

This feed measures, every day, how the Phionyx self-governance runtime governs the AI that writes Phionyx itself — aggregate numbers only, reviewer-reproducible. The headline is no longer just how often the gate ran (coverage); it is whether each governed decision was signed into a tamper-evident record and confirmed by a real outcome — the L3 lifecycle. Coverage is kept below as one input, not the goal.

L3 · the lifecycle, not just the invocation

Governed claims
159
Signed records
20
Real outcome labels
2

A governed decision is “complete” only when it is signed AND a real outcome confirms it — not merely when the gate ran. These three counts have different scopes — do not divide them. Per-claim completion is honestly near zero today because real outcomes accumulate over time; it is never fabricated. That honesty is the point.

Signed record chain
20 records · 1 chains

every governed decision is signed + hash-chained — tamper-evident and replayable from the records alone.

Invocation discipline · rolling 30-day window (one input, not the headline)

18.4%coverage · 432 commits · 159 gate calls (expected 864)

The CLAUDE.md rule expects two gate calls per commit (verify_claim + response_gate), so coverage measures how often the gate was invoked. At L3 this is kept as one falsifier, not the goal: coverage conflates a skipped gate with a commit made outside the harness, and a gate call never proves the work is correct — which is exactly why the lifecycle above (signed record + real outcome) is the headline.

Commits
432
in window
Gate calls
159
verify_claim + response_gate only
Expected
864
= 432 commits × 2
Sessions
49
MCP sessions in window

Per-day breakdown (27 days)

since 2026-05-04 17:48 UTC

DateCommitsGateAll MCPCoverage
2026-06-0323375.0%
2026-06-02112638118.2%
2026-06-0131114183.3%
2026-05-3127111.9%
2026-05-2919101026.3%
2026-05-2867515138.1%
2026-05-2724363675.0%
2026-05-2640152218.8%
2026-05-2534252.9%
2026-05-24264167.7%
2026-05-2327000.0%
2026-05-2217000.0%
2026-05-216000.0%
2026-05-203000.0%
2026-05-1918000.0%
2026-05-182000.0%
2026-05-1613000.0%
2026-05-1516000.0%
2026-05-131000.0%
2026-05-111000.0%
2026-05-1014000.0%
2026-05-094020.0%
2026-05-0810000.0%
2026-05-0718000.0%
2026-05-068000.0%
2026-05-0514000.0%
2026-05-047000.0%

Directive distribution

The aggregate of every directive written to telemetry across MCP tools + observability hooks — broader than the gate's five directive dispositions (pass · damp · rewrite · regenerate · reject) because non-gate tools also write entries.

auto_attest: 675pass: 163n/a: 16checkpoint: 8rewrite: 6reject: 2regenerate: 2require_tool: 1ok: 1

auto_attest — observability hooks (session start, user prompt log, commit attestation, …). Recorded but excluded from coverage math by design.

checkpoint — emitted by phionyx_checkpoint (lightweight context-note tool, not a gate).

n/a — entries without an explicit directive field (e.g. session start, prompt log, attestation writes).

solid / weak / incomplete — emitted by phionyx_causal_trace (chain-quality verdicts, distinct from the gate's five directive dispositions).

ok — emitted by the trust-boundary MCP server on successful third-party-tool attestation writes.

Reproduce

python3 case-studies/agentic-development-2026-05/scripts/runtime_evidence_self_audit.py --days 30

Same audit script lives at case-studies/.../scripts/runtime_evidence_self_audit.py. Generated 2026-06-03 14:48 UTC · schema v0.7.2 · daily refresh.

Agentic Development · Live governance feed — daily-refreshed