brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

Highlights

Nine new lifecycle skills let the agent handle bootstrapping, dependency audits, design reviews, performance analysis, refactoring, code review, scaffolding, and testing without manual invocation. A new ao harvest command pulls learnings from sibling workspaces so knowledge compounds across your entire multi-agent fleet, not just one repo. Context debugging is easier with ao context packet, and the hook system now formally supports both Claude Code and Codex runtimes.

What's New

• 9 lifecycle skills — bootstrap, deps, design, harvest, perf, refactor, review, scaffold, and test are now part of the RPI workflow with automatic invocation and mechanical gates
• Cross-rig knowledge harvesting — ao harvest extracts and catalogs learnings from sibling crew workspaces so insights travel between agents
• Context packet inspector — ao context packet lets you debug what inter-session handoff state the agent actually sees
• Dual-runtime hook support — Hooks now have a formal runtime contract covering Claude Code, Codex, and manual execution modes

All Changes

Added

• Nine lifecycle skills wired into the RPI workflow with auto-invocation
• Cross-rig knowledge consolidation via ao harvest
• Context packet inspection via ao context packet
• Hook runtime contract with Claude/Codex/manual event mapping
• Research provenance tracking on pending learnings
• Context declarations for inject, provenance, and rpi skills
• Evidence-backed output templates for goals and product commands

Changed

• Documentation reframed around three-gap context lifecycle model
• Hook docs updated with runtime modes table for dual-runtime support

Fixed

• Four pre-existing CI failures resolved
• Lookup retrieval gaps that caused empty results
• Embedded file sync on first session start
• Closure integrity with 24h grace window for evidence timing
• Skill lint compliance across vibe, post-mortem, crank, and plan
• Codex tool naming rule and five Claude-era tool references
• ASCII diagram consistency across 23 documentation files
• Fork exhaustion in validation script replaced with lightweight parser

Full changelog

Full Changelog: v2.30.0...v2.31.0

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.30.0 — Codex hookless lifecycle, PROGRAM.md workflows, and stronger long-running RPI runs

Highlights

AgentOps now handles Codex hookless sessions more cleanly, gives autonomous workflows a clearer PROGRAM.md contract, and makes long-running RPI runs much easier to inspect. This release also hardens the local release and validation path itself, so the same gate stack you rely on for shipping is more trustworthy under headless and generated-artifact-heavy workflows.

What's New

• Hookless Codex lifecycle support — Codex sessions can now run through startup, follow-up, validation, and closeout without depending on legacy hook assumptions.
• PROGRAM.md for autonomous work — Autodev and evolve flows now share a concrete program contract instead of relying on looser ad hoc context.
• Artifact-aware long RPI runs — Mission control now shows run artifacts and evaluator output so you can inspect what happened during multi-phase autonomous runs.
• More reliable release validation — Headless runtime checks, reverse-engineer hygiene, and release-gate coverage are more deterministic.

All Changes

Added

• Hookless Codex lifecycle support across CLI commands and skill orchestration
• A first-class PROGRAM.md contract for autodev and evolve-driven workflows
• Artifact and evaluator visibility for long-running RPI sessions

Changed

• Codex bundle maintenance, lifecycle guidance, and release validation coverage around the expanded Codex execution path

Fixed

• Codex RPI scope and closeout issues that caused follow-up and validation drift
• Release-gate regressions in headless runtime validation and learning coherence
• Reverse-engineer repo scans so generated or temporary trees no longer contaminate detected CLI surfaces

Full changelog

Full Changelog: v2.29.0...v2.30.0

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.29.0 — Config control, broader search, and stronger flywheel proof

Highlights

AgentOps now gives you more control over model spend, a broader default search path, and a tighter proof path for the knowledge flywheel. You can assign agent models by cost tier through config, ao search now pulls from both repo-local knowledge and upstream session history, and the flywheel claim is backed by deterministic proof fixtures instead of manual spot checks.

What's New

• Per-agent model routing — ao config now supports model cost tiers and direct config writes, so teams can tune quality and spend without manual file edits.
• Broader default search — ao search now brokers across upstream cass history and repo-local AgentOps artifacts instead of making you choose one surface up front.
• Stronger flywheel evidence — Close-loop validation now preserves research provenance and uses executable proof fixtures plus artifact-specific citation feedback.
• Richer review guidance — Council, research, swarm, vibe, athena, and post-mortem picked up new reference packs for reviewer routing, retrieval patterns, and write-time quality checks.

All Changes

Added

• Model cost tiers and direct config writes for per-agent routing
• Search brokerage across session history and repo-local knowledge
• New reference packs for reviewer routing, iterative retrieval, confidence scoring, conflict recovery, and write-time quality

Changed

• Comparison docs, command docs, and release smoke coverage around the expanded search and config surface

Fixed

• Flywheel proof, citation feedback, and closure reporting now agree on actual state
• Search stays aligned with forged session history and fallback behavior
• Pre-push and release validation is more deterministic under hook-launched git environments
• Council profile docs are synced between source and checked-in Codex artifacts

Full changelog

Full Changelog: v2.28.0...v2.29.0

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.28.0 — Competitive Feature Integration

Five features adopted from reverse-engineering GSD v1.27 and Compound Engineer v2.47:

Highlights

• Smarter failure recovery — Crank now classifies failures and auto-recovers (retry, decompose, or escalate) instead of blindly retrying
• Knowledge stays clean — Athena defrag runs at every session end, pruning stale artifacts automatically
• Per-project review config — Drop a .agents/reviewer-config.md to control which council judges run
• Right-sized plans — Plans auto-scale detail level (minimal/standard/deep) based on complexity
• Red-team your ideas — Brainstorm now stress-tests every approach before you choose

All Changes

See CHANGELOG.md for the complete list.

Full Changelog: v2.27.1...v2.28.0

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.27.1 — Hotfix: Flywheel golden signals now visible by default

The flywheel status was telling you everything was fine while hiding the full picture behind an opt-in flag. ao flywheel status said "COMPOUNDING" but the golden signals analysis (hidden behind --golden) said "accumulating." Now golden signals always compute and display — no more misleading status.

What changed

• Golden signals always shown — ao flywheel status now includes the four golden signals (velocity trend, citation pipeline, research closure, reuse concentration) and the overall verdict in every output format (table, JSON, YAML).
• --golden flag deprecated — Kept for backward compatibility but now a no-op (hidden from help).

Full changelog

See CHANGELOG.md for complete details.

Full Changelog: v2.27.0...v2.27.1

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

Highlights

The knowledge flywheel now tells you whether it's actually working. Four golden signals answer the question every agent operator asks: is my knowledge compounding, or just collecting dust?

ao flywheel status --golden

What's New

Golden Signals for Flywheel Health

Four health indicators that go beyond escape velocity (σρ > δ):

Each signal produces a verdict. Three or more healthy signals = compounding. Three or more critical = decaying. Mixed = accumulating — you know what to fix.

Forge-to-Pool Bridge

Forge now auto-writes pending learnings to .agents/knowledge/pending/ — closing the last manual gap in the flywheel loop. Knowledge flows from session → forge → pool → learnings → inject without intervention.

Session-Start Citation Priming

ao lookup runs at session start, surfacing relevant knowledge and creating the citation events that drive the feedback loop.

All Changes

Added

• Flywheel golden signals (ao flywheel status --golden)
• Forge-to-pool bridge for close-loop knowledge ingestion
• SessionStart citation priming via ao lookup
• Skill catalog quality improvements (descriptions, extraction, references)

Fixed

• .agents/.gitignore scope — replaced broad !*/ with explicit subdirectory list
• Codex runtime skill parity hardening
• Codex install smoke test assertions

Changed

• CLI reference docs regenerated

Full Changelog: v2.26.1...v2.27.0

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.26.1 — DAG-ify orchestrator skills

Hotfix: /rpi was stopping after implementation (Phase 2) without running validation (Phase 3). The execution steps were spread across prose sections with ### headings that created natural LLM stopping points.

Highlights

• RPI now runs all three phases reliably. The execution sequence for /rpi, /discovery, and /validation is encoded as a compact DAG code block — no section breaks between steps, no natural stopping points for the LLM.
• -577 lines across 6 skill files (3 source + 3 codex variants). Less prose, more program.

What's New

Fixed

• /rpi stops after Phase 2 — restructured as compact DAG
• /discovery and /validation restructured to match
• Test patterns updated for new heading format

Changed

• GOALS.md rebuilt from first principles
• README leads with moats, progressive disclosure
• CLI reference docs regenerated
• Doctor + findings helper test coverage added

Full Changelog: v2.26.0...v2.26.1

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.26.0 Release Notes

Highlights

• Test pyramid expanded to BF1–BF9 — Four new bug-finding levels cover regression replay, performance benchmarks, backward compatibility, and security-in-test patterns
• Language-specific test patterns — Go and Python standards now include concrete examples for every new BF level
• Codex audit: 60+ fixes — Orphaned references removed, lint warnings resolved, manifest hashes regenerated across all 54 Codex skills

What's New

• BF6 (Regression): Bug-specific replay tests with ID-based naming (TestBug_AG_XYZ_... / test_bug_ag_xyz_...)
• BF7 (Performance): Benchmark patterns using Go testing.B and Python pytest-benchmark
• BF8 (Backward Compatibility): Fixture corpus approach with testdata/compat/ (Go) and tests/fixtures/compat/ (Python)
• BF9 (Security): In-test secrets redaction and path traversal rejection patterns
• Decision tree extended with 4 new routing questions
• RPI phase mapping updated: bug fix mandates BF6, hot-path mandates BF7, format changes mandate BF8, secrets mandate BF9
• regen-codex-hashes.sh script for Codex manifest maintenance

All Changes

Full Changelog: See CHANGELOG.md

Full Changelog: v2.25.1...v2.26.0

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

Fixed

• Codex BF pyramid parity — Synced BF1/BF2/BF4 bug-finding level selection into skills-codex implement, post-mortem, and validation skills
• Codex Claude backend cross-contamination — Removed orphaned backend-claude-teams.md files (Claude primitives: TeamCreate, SendMessage) from 4 Codex skills (council, research, shared, swarm)
• Dead converter rule — Removed stale sed substitution for backend-claude-teams.md rename in converter script
• Swarm reference integrity — Added Reference Documents section to swarm SKILL.md; updated validate.sh to check only Codex-native backend references

Full Changelog: v2.25.0...v2.25.1

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance

v2.25.0 — Test Pyramid + Autonomous Execution

Highlights

• Test pyramid baked into every RPI phase. L0–L7 levels flow from discovery through post-mortem. Agents own L0–L3 autonomously; L4+ requires human input. Plans include test level metadata, pre-mortem validates coverage, post-mortem reports gaps.

• RPI and Evolve now run fully autonomous by default. No human questions between phases. Three-Phase Rule enforces discovery → implementation → validation as a single uninterrupted flow. Anti-pattern tables catch 13 common failure modes.

• Codex skill infrastructure matures. New API contract, DAG-based smoke test for 54 skills, durable overrides for crank/swarm/council, and a complete standards reference for Codex skill authoring.

What's New

Test Pyramid Standard

A shared reference (test-pyramid.md) defines 8 test levels with clear agent autonomy boundaries. Every RPI phase now knows which test levels to scope, plan, write, and validate.

Autonomous Execution

/rpi and /evolve enforce hands-free execution. No pausing to ask, no stopping after implementation, no narrating plans. The human touchpoint is the final report after all three phases complete.

Codex Platform

Output contracts on verdict skills, a conformance validator, and converter improvements that properly strip Claude primitives instead of mapping to non-existent tools.

Full Changelog: v2.24.0...v2.25.0