v0.2.0

What's Changed

• S107: tagged binary release v0.1.0 — the no-Rust install path by @ifelse-codes in #112
• S108: publish to crates.io + Homebrew tap — finish every install channel by @ifelse-codes in #113
• S108 follow-up: S109 first slice = real paid Researcher sub-agent by @ifelse-codes in #114
• S109: fleet slice 1 — Researcher as a governed Claude Code subagent by @ifelse-codes in #115
• S110: NO-CODE Ground Truth (audits S106-S109) — PARTIAL by @ifelse-codes in #116
• S111: close the fleet's def-vs-dispatch wire by @ifelse-codes in #117
• S112 — CODE: downstream handoff-consumption (the fleet's findings reach the pipeline) by @ifelse-codes in #118
• S112 post-merge: record the merge in the spine by @ifelse-codes in #119
• S113 — fleet work visible to the counter (beside K), second role chosen by @ifelse-codes in #120
• S113 post-merge: record the merge in the spine by @ifelse-codes in #121
• S114: the fleet's second role — the Fidelity Reviewer by @ifelse-codes in #122
• S114 post-merge: record the merge in the spine + downgrade an S111 claim by @ifelse-codes in #123
• S115: NO-CODE Ground Truth (audits S111-S114) — PARTIAL PASS by @ifelse-codes in #124
• S116: the fleet's third role, the Plan Advisor by @ifelse-codes in #125
• S117: prove the Plan Advisor dispatches by name by @ifelse-codes in #126
• S118 — Dogfood (paid): the overdue vajra claude run found a green gate over a broken build by @ifelse-codes in #127
• S118 closeout: re-scope S119 to the clean-room re-run + record the no-release directive by @ifelse-codes in #128
• S119: the clean-room runner (QA + Demo-er run in an environment the agent did not prepare) by @ifelse-codes in #129
• S120 GT: PARTIAL PASS + S121 qa-specialist prompt by @ifelse-codes in #130
• Session 121 — the QA Specialist: the fleet's 4th role, the first that executes by @ifelse-codes in #131
• S121 closeout: record PR #131 as MERGED (stale-open-PR drift) by @ifelse-codes in #132
• S122 — close the four real holes the live QA run found in S121's guardrails by @ifelse-codes in #133
• S122 closeout: record PR #133 MERGED by @ifelse-codes in #134
• S122 closeout fix: correct a wrong dogfood-staleness claim by @ifelse-codes in #135
• S123 prompt: founder picks the clean room as the fence by @ifelse-codes in #136
• S123: fence the qa-specialist Write/Edit grant by @ifelse-codes in #137
• S124 prompt: swap the receipt session for a real dogfood run by @ifelse-codes in #138
• S124: paid dogfood — does the fleet + fence machinery hold under real use? by @ifelse-codes in #139
• S125: NO-CODE ground truth + full-stack review — PARTIAL PASS by @ifelse-codes in #140
• S126: finish the SDLC agent fleet — the last five roles (roster complete at nine) by @ifelse-codes in #141
• S127 prompt, rewritten: every recommendation must be ANSWERED (obeyed / refused / deferred) by @ifelse-codes in #143
• S127: every recommendation must be ANSWERED — the first gate that consumes a handoff by @ifelse-codes in #144
• S127 post-close: founder picked C — first contact works by @ifelse-codes in #145
• S128 — CODE: first contact works by @ifelse-codes in #146
• S128 closeout — founder pick recorded: A by @ifelse-codes in #147
• S129: one source for what a stranger gets by @ifelse-codes in #148
• S130: mandatory NO-CODE ground truth (audits S126-S129) by @ifelse-codes in #149
• S131: fidelity-reviewer handoff mandatory, provenance provable by @ifelse-codes in #150
• S131 hotfix: satisfy clippy::unnecessary_sort_by (CI red on main) by @ifelse-codes in #151
• S132: an obeyed: disposition must be judged true, not merely resolve by @ifelse-codes in #152
• S132 closeout amendment: re-sequence S133 to the design-advisor mandate by @ifelse-codes in #153
• S133 — the design-advisor becomes mandatory, and a skip must carry a recorded reason by @ifelse-codes in #154
• S133 closeout amendment — S134 = the paid dogfood on chitra, S135 = implementation-advisor (locked) by @ifelse-codes in #155
• S134: PAID DOGFOOD — a real governed session in chitra, and the brownfield threshold hole it found by @ifelse-codes in #156
• S134 closeout amendment: S135 rescoped for a $20/month plan; chitra deferred by @ifelse-codes in #157
• S134 closeout: persist the three carry-forward rules into .ai/ by @ifelse-codes in #158
• S135: the tech-lead — the tenth role, and the binding --check-crew gate (ACCEPT 11/12) by @ifelse-codes in #159
• S136: vajra init --sync-fleet — the upgrade path skip-if-present could never be, and the fleet made real in chitra (ACCEPT 6/9) by @ifelse-codes in #160
• S136 hotfix: cargo fmt — CI red on main, a recurrence of S96 by @ifelse-codes in #161
• S136 follow-up: check 9 asserted the wrong thing once chitra committed the roster by @ifelse-codes in #162
• S136 follow-up: the S137 draft carried a guardrail that S136 inverted by @ifelse-codes in #163
• S136 follow-up: approve the S137 prompt, and correct STATE to post-close reality by @ifelse-codes in #164
• S137: PAID DOGFOOD — chitra's scatter chart locked to the reference language (ACCEPT 5/5) by @ifelse-codes in #165
• S137 follow-up: correct the record — the 'blind spot' was a setup artifact, not a Vajra failure by @ifelse-codes in #166
• S138 — THE REAL DOGFOOD: vajra claude run from INSIDE chitra (heatmap lock) by @ifelse-codes in #167
• S139 — make required bind at CLOSE: check_required_crew in verify-closeout.sh by @ifelse-codes in #168
• S140 — mandatory NO-CODE Ground Truth closeout (🟡 PARTIAL PASS) by @ifelse-codes in #169
• S141: best install + upgrade-in-place — recorded render provenance by @ifelse-codes in #170
• S142: complete the upgrade loop — the render stamp generalises to shell hooks by @ifelse-codes in #171
• S143: the constitution joins the smooth upgrade (split fill from governed body) by @ifelse-codes in #172
• S144: the chitra FULL-LOOP dogfood — upgrade + govern a build to green close by @ifelse-codes in #173
• S145: mandatory NO-CODE Ground Truth — 🟡 PARTIAL PASS by @ifelse-codes in #174
• S146: propagate verify-closeout.sh via --sync-fleet + PATH-first resolver by @ifelse-codes in #175
• S146 closeout: SESSION→147, SESSION-BOOT, TASK by @ifelse-codes in #176
• S147: prove the 5 quiet fleet roles (all 5 Changed) by @ifelse-codes in #177
• S148: close test-runner compression gaps (bare jest + fail-path 30-399 lines) by @ifelse-codes in #178
• S149: advice-influence audit — F2f gap measured by @ifelse-codes in #179
• S150 GT: 🟡 PARTIAL PASS — cargo fmt 🔴, pipeline 3-5/8, F2f 59% by @ifelse-codes in #180
• S151: fix cargo fmt + add fmt check to verify-closeout.sh by @ifelse-codes in #181
• S152: obedience-skip + carry-forward rules in AGENTS.md by @ifelse-codes in #182
• S152: closeout pointer — TASK → S153, S153 prompt by @ifelse-codes in #183
• S153: close 4 carry-forward items (AGENTS.md prose + cargo threshold test) by @ifelse-codes in #184
• S154: CODER station — tighten execution-sha guard (step-sha traces) by @ifelse-codes in https://github.com/ifelse-code...

v0.1.0

What's Changed

• feat: installer / release path (S13) by @ifelse-codes in #1
• feat: maturity levels L1/L2/L3 (session 14) by @ifelse-codes in #2
• docs: S15 ground truth audit + closeout by @ifelse-codes in #3
• S16: remove vajra launch alias + fix S15 drift by @ifelse-codes in #4
• S16 closeout: state sync + S17 prompt by @ifelse-codes in #5
• Add vajra estimate command for pre-run cost prediction by @ifelse-codes in #6
• Session 17 closeout by @ifelse-codes in #7
• Session 18 — Product review + Varta direction by @ifelse-codes in #8
• S19: Varta v0 — the ⚡ language (skill) by @ifelse-codes in #9
• S20: ground-truth now catches direction drift (vision + roadmap) by @ifelse-codes in #10
• S21: co-pilot loader — Varta's ⚡on(cond) ⚡include fires (and enforces) by @ifelse-codes in #11
• S22: scaffold propagation — vajra init emits S20 GT audits + S21 co-pilot by @ifelse-codes in #12
• S23: first-run "aha" — vajra init ends with a live co-pilot fire by @ifelse-codes in #13
• S23 enforcement: one-session-per-chat → Hard Rule + roadmap Vajra enforcement by @ifelse-codes in #14
• S24: render .ai/ → generated vajra.varta (drift-guarded) by @ifelse-codes in #15
• S25 ground-truth closeout (NO-CODE) by @ifelse-codes in #16
• S26: enforce one-session-per-chat (session-guard hook) by @ifelse-codes in #17
• S27: Darshan — the human's glanceable output skill by @ifelse-codes in #18
• S28: propagate Darshan into vajra init by @ifelse-codes in #19
• S28 closeout bundle by @ifelse-codes in #20
• S29: propagate one-session-per-chat guard into vajra init by @ifelse-codes in #21
• S30 ground-truth: founder-satisfaction gate + dogfood_check hardening by @ifelse-codes in #22
• S31 — dogfood/verification: gate verdict + 3 core findings (docs-only) by @ifelse-codes in #23
• S32: Darshan enforcement — surface it in the boot packet (advised → enforced) by @ifelse-codes in #24
• docs: capture headroom discovery — detector vs enforcer + obedience ladder by @ifelse-codes in #25
• docs: surface the deferred S33 build-order fork at boot by @ifelse-codes in #26
• S33: compression schema fix — HookInput drops camelCase rename (S31 #2) by @ifelse-codes in #27
• docs: fold claude-mem refinements into obedience/pace-notes backlog by @ifelse-codes in #28
• S34: Brownfield onboarding — session-0 brief, .ai/hooks placement, auth pre-check by @ifelse-codes in #29
• S35 closeout: ground-truth gate re-measure (NO-CODE) by @ifelse-codes in #30
• S36: real dogfood run — enforcement moat leaks + gate re-measure by @ifelse-codes in #31
• S37: guard outward/irreversible actions (close the S36 enforcement leak) by @ifelse-codes in #32
• S38: propagate the publish-guard into vajra init by @ifelse-codes in #33
• session 39 harden guards by @ifelse-codes in #34
• S40: ground-truth (NO-CODE) — enforcement-completeness audit by @ifelse-codes in #35
• S41: fix the compression fail-gate (correctness-first) by @ifelse-codes in #36
• S42: jq-preflight fail-closed on all 5 hooks (Gap 1, founder pick C) by @ifelse-codes in #37
• S43: scaffold git-level belt (.githooks/pre-commit + pre-push) into vajra init by @ifelse-codes in #38
• S44: merge into existing .claude/settings.json on vajra init by @ifelse-codes in #39
• S45: combined 3-lens ground-truth by @ifelse-codes in #40
• session 46 live redogfood by @ifelse-codes in #41
• S47: mid-run co-pilot murmur + disable publish-guard (founder directive) by @ifelse-codes in #42
• S48: obedience metric — clean/blocked/obedience% from the session trace by @ifelse-codes in #43
• S49: obedience baseline — vajra meter --all (batch the S48 metric) by @ifelse-codes in #44
• S50: Ground Truth (NO-CODE) — dogfood/enforcement audit + S51 closeout by @ifelse-codes in #45
• S51 prompt: retarget value-gap test to the real chitra repo by @ifelse-codes in #46
• S51 — Measure the value gap (real-task A/B on chitra, PAID) by @ifelse-codes in #47
• S52 — Value gap on a HARDER task (n=2 null) + reframe to governance (S53) by @ifelse-codes in #48
• S53 — Reframe to governance, refined to a governed multi-agent SDLC pipeline (NO-CODE) by @ifelse-codes in #49
• S53 fix: Analyst generates the next PROMPT (not a spec.md) + 'map concepts to Vajra' learning by @ifelse-codes in #50
• S54: the Analyst stage + fidelity-over-discipline pivot (DECISION-002) by @ifelse-codes in #51
• S55: fidelity auditor brain — cold re-audit of S54 (NO-CODE ground-truth) by @ifelse-codes in #52
• S56: the fidelity GATE (teeth) — closeout structurally requires an independent ACCEPT by @ifelse-codes in #53
• S57: propagate the fidelity gate + reviewer into vajra init by @ifelse-codes in #54
• S58: verdict-authorship attestation — bind an ACCEPT to the delivery it reviewed by @ifelse-codes in #55
• S59: the attested-verdict delta ledger (the headline moat's first code) by @ifelse-codes in #56
• S60 — mandatory NO-CODE ground-truth (lens A: partial scope-creep → pivot to payload) by @ifelse-codes in #57
• S61: Analyst Generate + Delta half made REAL (S54 REJECT 1-of-5 → 3-of-5) by @ifelse-codes in #58
• S62 — Analyst: Intake (J1) + Options (J2) — close the S54 REJECT (3-of-5 → 5-of-5) by @ifelse-codes in #59
• S63 — Paid dogfood: the governed loop measured as experience (ACCEPT) by @ifelse-codes in #60
• S64 — The Planner stage (pipeline station 2): coverage-checked plan gate by @ifelse-codes in #61
• S65 — Ground Truth (mandatory NO-CODE, every 5th): 8 audits + S66 pick B by @ifelse-codes in #62
• S66: make the receipt authoritative (retire the ~4.71× overstatement) by @ifelse-codes in #63
• S67: the Architect stage — the pipeline's DESIGN gate (4th governed station) by @ifelse-codes in #64
• S68: the Coder handoff — the pipeline's CODE/execution gate (the LAST station) by @ifelse-codes in #65
• S69: the QA station — verification becomes a live-executed close gate (pipeline station 6) by @ifelse-codes in #66
• S70 — Ground Truth (mandatory NO-CODE, every 5th): 8 audits, lens-A PARTIAL PASS, founder decisions + S71 Demo-er pick by @ifelse-codes in #67
• S71: the Demo-er station — pipeline station 7, the SHOW gate by @ifelse-codes in #68
• S72: the Releaser station — pipeline station 8, the SHIP gate by @ifelse-codes in #69
• S73 re-pick: close-path reliability (park the dogfood ride-along) by @ifelse-codes in #70
• S73 — Close-path reliability: fix the brakes (deflake + bounded live gate runs) by @ifelse-codes in #71
• S74: the payload counter — measure whether the pipeline advances by @ifelse-codes in #72
• S75 GT: pipeline-advance counter's first live reading (1/8->8/8) + 2 new findings by @ifelse-codes in #73
• S76 — Dogfood ride-along (paid MEASURE): the 8-station pipeline as lived experience by @ifelse-codes in #74
• S77 — Receipt truth on real runs: price fable-5, stop faking a total by @ifelse-codes in #75
• S78 — Recover the true $: capture Claude Code's own end-of-...