c6ce59e

What's Changed

• chore(deps): update Composer dependencies by @Korbeil in #17
• refactor: remove duplicated tracker, permission, number and path helpers by @Korbeil in #18
• feat(poller): park tasks whose PR was closed in a new closed state by @Korbeil in #19
• docs(config): mark sync/poll/review/ci keys as required, not defaulted by @Korbeil in #20
• fix(store): check record writes and skip unreadable records by @Korbeil in #21
• fix(gh): read every review and thread, not the oldest 100 by @Korbeil in #22
• fix: name /pablo-commit-and-pr in user-facing messages by @Korbeil in #24
• fix(linear): bound every linear CLI call with a 30 s timeout by @Korbeil in #25
• chore(castor): let .php-cs-fixer.php alone list the fixed paths by @Korbeil in #26
• fix(web): only answer dashboard requests for trusted hosts (loopback + dashboard_trusted_hosts) by @Korbeil in #23
• fix(agents): detach background runs without setsid on macOS by @Korbeil in #27
• fix(sync): isolate per-worktree failures and skip closed tasks by @Korbeil in #28
• fix(git): give git calls explicit timeouts and surface them as PabloError by @Korbeil in #29
• fix(startup-script): watch the script to completion and run it from the worktree by @Korbeil in #31
• ci: harden the CI workflow and test the lowest supported PHP by @Korbeil in #32
• chore(composer): declare the PHP extensions the code uses by @Korbeil in #35
• docs(readme): add screenshot, safety summary, full requirements and links by @Korbeil in #36
• refactor(paths): resolve HOME and ~ expansion only through PabloPaths by @Korbeil in #37
• feat(sync): add --dry-run to sync:run and sync:main by @Korbeil in #38
• refactor(di): require process runners and usage harvesters in constructors by @Korbeil in #45
• feat(update): keep the replaced binary and add pablo update --rollback by @Korbeil in #46
• feat(cli): target any task by branch and filter task:list by @Korbeil in #49
• feat(review): count whitelisted bot reviews through their unresolved threads by @Korbeil in #57
• feat(analytics): run breakdowns, task history and per-model views in the CLI and on /analytics by @Korbeil in #50
• docs(site): make the safety claims exact by @Korbeil in #30
• fix(config): skip an invalid project file instead of failing every project by @Korbeil in #33
• feat(dispatch): record dispatcher health, back off failing jobs by @Korbeil in #55
• test: stop the suite waiting out sleeps and reaching a real launcher by @Korbeil in #42
• fix(sync): leave an unpushed local rewrite alone by @Korbeil in #34
• fix(install): compute the services' PATH at install time by @Korbeil in #40
• feat(cli): add --format=table|json to the read commands by @Korbeil in #43
• docs(agents): refresh AGENTS.md against the current code by @Korbeil in #52
• fix(gh): don't count cancelled or superseded check runs as CI failures by @Korbeil in #58
• docs: gather runtimes and ADEs on one page, split update out of installation by @Korbeil in #47
• docs: one name for the agent environment (ADE), and a glossary by @Korbeil in #44
• docs: tighten wording, use real relaunch numbers, unify terms by @Korbeil in #41
• feat(sync): keep a per-project sync history by @Korbeil in #51
• feat(hooks): run the user's own command on transitions and finished agents by @Korbeil in #54
• feat(prune): add system:prune, run daily at 03:00 by its own timer by @Korbeil in #48
• docs: fix small mismatches between the docs and the code by @Korbeil in #39
• feat(cli): complete task branches, projects and states, install the completion by @Korbeil in #53
• test: boot every test kernel in debug so a stale container is rebuilt by @Korbeil in #62
• refactor(listing): split the task listing into gather and render steps by @Korbeil in #63
• feat(config): validate config keys and values, add system:config:validate by @Korbeil in #64
• site: fix accessibility gaps, self-host fonts, add sitemap and canonicals by @Korbeil in #61
• docs(quickstart): show real output after each step, add a model step by @Korbeil in #60
• refactor(json): decode JSON through a checked Json helper by @Korbeil in #59
• refactor(errors): checked Files helpers instead of bare @ file calls by @Korbeil in #65
• refactor(config): group ProjectConfig into per-section value objects by @Korbeil in #66
• refactor(config): make ConfigSchema the only validator by @Korbeil in #68
• fix(prune): check pid liveness without requiring ext-posix by @Korbeil in #67
• refactor(storage): share one file lock, atomic write and stamp writer by @Korbeil in #69
• refactor(runtime): share agent translation, freshness and usage window helpers by @Korbeil in #71
• fix(sync): push task branches to their own remote in a fork setup by @Korbeil in #72
• refactor: delete code with no production caller by @Korbeil in #70
• refactor(doctor): one probe map, shared registry and ADE bases by @Korbeil in #77
• refactor(state-machine): one launch table and a single launch() by @Korbeil in #74
• refactor(agents): drop the watcher's unused --then/--expect-state options by @Korbeil in #73
• test: build the fake graph in one place and make temp dirs collision-proof by @Korbeil in #75
• refactor(opencode): assign session ids up front instead of looking them up by @Korbeil in #78
• refactor(di): inject the ADE/runtime registries and Hooks by @Korbeil in #83
• refactor(gh): one reviewVerdict() call and a shared gh pr list helper by @Korbeil in #79
• refactor(cli): share the analytics loading, project selection and output helpers by @Korbeil in #81
• refactor(agents): derive launcher session queries and slim constructors by @Korbeil in #80
• refactor(cli): inject only what each command uses by @Korbeil in #82

Full Changelog: c5abd3c...c6ce59e

171a914

What's Changed

• test: remove every TempDir path when the test process exits by @Korbeil in #84
• perf(dispatch): poll projects concurrently, bounded and rate-limit aware by @Korbeil in #85
• perf(poller): fewer gh/git subprocesses per poll (dedupe + one GraphQL query per repo) by @Korbeil in #89
• test: run the suite in parallel with paratest by @Korbeil in #88
• perf(cli): load commands lazily by @Korbeil in #87
• perf(tests): build git fixtures once per process, copy them per test by @Korbeil in #90
• perf(dispatch): skip the CLI preflight on ticks with nothing due by @Korbeil in #86

Full Changelog: c6ce59e...171a914

e01bd19

What's Changed

• feat(claude-code): run headless agents as Remote Control sessions by @Korbeil in #15

Full Changelog: f1e9efe...e01bd19

c5abd3c

What's Changed

• fix(ade): make OpenCode agents reach Orca and OpenChamber by @Korbeil in #16

Full Changelog: e01bd19...c5abd3c

f1e9efe

What's Changed

• fix(ci): make phpstan and tests green on GitHub Actions by @Korbeil in #2
• fix(store): flatten slashes in branch names to state filenames by @Korbeil in #3
• feat(state-machine): approved state + testing.enabled to disable the QA flow by @Korbeil in #4
• feat(agents): per-project-type agent model resolution by @Korbeil in #5
• chore(deps): update dependencies and CI actions by @Korbeil in #6
• feat(site): add GitHub Pages website and new logo by @Korbeil in #7
• feat(agents): run the agents with Claude Code (agent_runtime: claude-code) by @Korbeil in #8
• feat(agents): run the agents with GitHub Copilot CLI (agent_runtime: copilot) by @Korbeil in #9
• feat(install): interactive curl | bash installer at getpablo.dev/installer by @Korbeil in #10
• feat: distribute PABLO as a standalone binary by @Korbeil in #11
• refactor(runtime): confine runtime-specific code to src/Agents/Runtime// by @Korbeil in #13
• refactor(ade): confine ADE-specific code to src/Agents/Ade//, headless as its own ADE by @Korbeil in #14

Full Changelog: https://github.com/Korbeil/pablo/commits/f1e9efe