v0.4.5

Added

• config/profile-resolution.md — the shared profile-resolution rule now lives in one file that skills reference, instead of being restated inline in every skill. job-tracker:import and job-tracker:run keep their own profile logic.
• migrations/0.4.5.md — registers config/profile-resolution.md in the config/paths.md zone list, and replaces the now-superseded inline profile-resolution wording in existing config/settings.md and config/tracker-schema.md with a delegation to config/profile-resolution.md. The new file itself is delivered by npx llm-job-tracker update (managed entry) or a plugin reinstall, not by this migration.

Changed

• Skill Next actions footers now always print as plain Markdown (letter shortcuts + commands), and additionally call AskUserQuestion for a clickable choice when that tool is available in the current session — plugin/npx installs running in hosts without AskUserQuestion (e.g. Codex) fall back to the plain-text footer instead of failing. Shortcut/free-text mapping, freshness, and confirm-before-acting rules are centralized in config/next-actions.md. AskUserQuestion options are capped at 4 per question (was documented as up to 5).
• config/settings.md's ## Profile Rules and config/tracker-schema.md's ## Update Rules no longer restate the profile-resolution logic; both defer to config/profile-resolution.md. job-tracker:run now reads config/profile-resolution.md for per-stage resolution instead of config/settings.md.
• config/profile-resolution.md now specifies how to find the "matching" tracked row: by URL when available, otherwise Company + Role, asking the user to disambiguate on multiple matches.

Fixed

• node scripts/tracker.js move --from raw --to active no longer throws cannot preserve columns when the Raw Pipeline row has Added/Source values and the Active Pipeline table has no Notes/Detail column — those two fields are intake-only metadata and are now dropped silently instead of blocking the move. Other genuinely unmapped columns still raise the error.
• That same drop no longer applies when the destination table does have a Notes/Detail column (e.g. raw → archive/monitoring): Added/Source are stashed there as before. The intake-metadata drop only kicks in as a last resort, when there is truly nowhere to preserve the value.

v0.4.4

Added

• node scripts/tracker.js add-lead --url <url> now auto-derives the Source column from config/source-registry.md — no need to pass --source explicitly when adding a lead from a known job board URL.
• migrations/0.4.4.md — registers config/source-registry.md and config/browser-patterns.md in the config/paths.md zone list, and updates config/next-actions.md example footers to the slash-command form.

Changed

• node scripts/ats-probe.js derive-source <url> resolves the correct tracker Source value for any job URL using your workspace config/source-registry.md. Falls back to the bare domain when no pattern matches. Useful for auditing source labels on existing tracker rows.
• job-tracker:health browser-required source check is now driven by your config/source-registry.md instead of a hardcoded list — adding a new browser-required source to the registry is enough; no script update needed. Each browser-required source must have a matching Source Derivation entry; missing entries are reported as errors (previously only Djinni was checked).
• Scripts reorganized into focused modules: scripts/ats/, scripts/tracker/, scripts/workspace/, scripts/publish/, scripts/deps/. Thin entry-point wrappers (scripts/ats-probe.js, scripts/tracker.js, scripts/check-deps.js, etc.) are preserved for backwards compatibility.

Fixed

• ats-probe.js profile keyword loading now resolves relative to the detected workspace root instead of process.cwd(), so profile hints apply correctly when the CLI is run from a subdirectory.
• ats-probe.js, tracker.js, and check-public.js entry-point detection now resolves symlinks before comparing paths, so the CLI runs correctly when invoked through a symlinked directory (e.g. macOS /tmp -> /private/tmp) instead of silently exiting with no output.
• ats-probe.js discover now reports a provider whose every slug candidate failed as an error entry (count: 0, error: <reason>) instead of silently dropping it; text output prints the error reason instead of a bare zero-count row.
• job-tracker:health now requires Playwright MCP with the user's logged-in account/session for every login-gated browser-required source (detected from its "Why browser is required" column), not just a hardcoded LinkedIn/Djinni pair — closing a gap where a login-gated source without the word "Playwright" in its policy passed silently.

v0.4.3

Added

• config/source-registry.md — source values, host patterns, ATS probe providers, browser-required sources such as Djinni/LinkedIn, and URL-to-Source derivation rules.
• migrations/0.4.3.md — seeds config/source-registry.md for existing workspaces that already completed 0.4.2, and adds source-registry references to protected strategy/sources.md and config/tracker-schema.md.
• Health checks for source-registry drift: scripts/check-workspace.js now verifies that configured ATS providers match scripts/ats-probe.js and that Djinni has both browser-required and source-derivation policy.

Changed

• job-tracker:find, job-tracker:import, and job-tracker:verify read config/source-registry.md instead of hardcoding provider/source policy in skill files.
• scripts/ats-probe.js now reads provider IDs, discovery feed templates, default keywords, and default locations from config/source-registry.md; JavaScript keeps only provider-specific payload normalization.
• strategy/sources.md, config/tracker-schema.md, config/browser-patterns.md, and config/agent-instructions.md now delegate source values and browser-required source policy to config/source-registry.md.
• job-tracker:verify liveness strategy clarified: Browser MCP on the direct job URL is the primary liveness check for tracked roles; scripts/ats-probe.js and ATS board APIs are discovery surfaces only and must not be used as the closing signal for a specific tracked job ID.

v0.4.2

Added

• scripts/ats-probe.js — reusable Node.js ATS probe for Ashby, Lever, Greenhouse, Workable, Recruitee, and SmartRecruiters. It normalizes role title/location/id/URL, filters frontend/product/fullstack/platform leads, and marks duplicates against data/tracker.md.
• scripts/tracker.js — structured Markdown tracker CLI for listing rows, adding Raw Pipeline leads, moving rows between sections, setting status, and bumping date fields without brittle text replacement.
• scripts/ats-probe.js discover <company-or-domain> — deterministic ATS slug discovery by probing supported provider APIs with normalized company/domain candidates.
• scripts/tracker.js validate, --dry-run, --json, and section-scoped row updates.
• Tracker CLI schema aliases for English and Ukrainian section/field labels, documented in config/tracker-schema.md.
• scripts/tracker.js exports setSchemaRootFromTracker() for programmatic callers that run outside the workspace cwd.
• config/browser-patterns.md — Browser MCP operational guide for stable evaluate calls, locator-based clicks, cookie overlays, login-required sources, and browser safety boundaries.
• migrations/0.4.2.md — seeds Browser MCP patterns and adds ATS probe/browser guidance to protected existing workspaces.
• Tests for ATS normalization/filtering/dedupe/discovery, tracker row operations/CLI flags, and run stop-check warnings.
• README.md — ## Agent CLI Utilities section documenting ATS probe and tracker CLI commands, options, and usage patterns for agents and maintainers.

Changed

• job-tracker:find, job-tracker:import, and job-tracker:verify now prefer scripts/ats-probe.js for supported ATS sources before ad hoc API parsing or browser fallback.
• job-tracker:find and job-tracker:import read config/browser-patterns.md at startup.
• job-tracker:verify reads config/browser-patterns.md at startup and prefers scripts/ats-probe.js for supported ATS liveness checks before browser fallback.
• job-tracker:run now explicitly prefers scripts/tracker.js for tracker row updates and requires profile switching through job-tracker:profile use.
• job-tracker:run completion rules now forbid Status: done while internal queue work remains, background subagents are still running, or skipped selected leads lack tracker-recorded real reasons.
• job-tracker:run orchestrator enforces "tool call before text" as a hard rule when state is running; a text-only response is invalid unless state is paused-resumable, blocked, or done. Next internal step: written without an accompanying tool call is now documented as a stop-point anti-pattern.
• job-tracker:fit documents parallel batch fit-review behavior for job-tracker:run when subagents are available.
• strategy/sources.md documents canonical scripts/ats-probe.js probe commands in the ATS section as the preferred alternative to ad hoc curl/jq parsing.
• Stop hook now warns when a job-tracker:run response reports done without completion-guard evidence or mentions background/subagent work without paused-resumable, Continue Run, or Next internal step:.
• Stop hook generic reminders now only appear for job-tracker-like outputs, and run-done detection is scoped to job-tracker:run context.
• scripts/llm-hooks/validate-skill-footers.js now checks skills/run/SKILL.md for required run completion guards (internal action queue is empty, no background subagent, every skipped selected lead has a reason recorded, never \done`, Continue Run`).
• scripts/llm-hooks/codex-rules/default.rules uses npm run validate:skill-footers instead of the direct node invocation.
• scripts/check-workspace.js now loads tracker section and field aliases from config/tracker-schema.md before parsing data/tracker.md, enabling localized or custom column labels without code changes.

Fixed

• Run orchestration now documents paused-resumable as the required visible state when a turn must end while background subagent work can continue automatically.
• Angular/non-React stack assumptions are explicitly forbidden as invented skip reasons unless the active profile says so.

v0.4.1

Fixed

• Preserve legacy scripts/resume.css during workspace update so the next migration can move user-customized PDF CSS to scripts/cv.css instead of losing it before job-tracker:health runs.
• migrations/0.4.1.md migrates existing workspaces from scripts/resume.css to scripts/cv.css and updates the protected config/paths.md reference.

v0.4.0

Added

• job-tracker:import <url> — new command: imports one job posting URL into Raw Pipeline. Verifies the posting at source (browser MCP for JS-rendered/login pages), auto-selects the best-fit profile across all configured profiles and switches to it (asking on tie), derives Source from the URL host (ashby/greenhouse/lever/linkedin/workable/etc.), and writes a per-import session report at .sessions/reports/[id].import.md. Deduplicates against existing tracker URLs and best-effort recent import reports.
• Fit Rubric — strategy/criteria.md now contains ## Fit Rubric: a three-layer model (Hard Gates PASS/DISQUALIFIED/UNRESOLVED → six-dimension /60 fit score → Application ROI modifier), six dimensions with 1/3/5 anchors, and a per-dimension evidence matrix with Unknown → 3, never fabricate rule. skills/fit/SKILL.md updated to apply the rubric with Hard Gates before scoring and ROI as a recommendation modifier only.
• Session report filename generalized to [id].<skill>.md ([id].run.md for run, [id].import.md for import). check-workspace now accepts any kebab-case skill suffix; blocked import reports are not flagged as unfinished runs.
• migrations/0.4.0.md — three idempotent migration steps for existing workspaces: (1) add ## Fit Rubric to strategy/criteria.md, (2) update session-report path in config/paths.md to [id].<skill>.md, (3) update profile-switch exception in config/settings.md to name both job-tracker:run and job-tracker:import.

Changed

• job-tracker:import is now the second skill (besides job-tracker:run) allowed to switch the active profile via job-tracker:profile use. Profile switch rules updated in skills/profile/SKILL.md, config/settings.md, README.md, and config/agent-instructions.md to distinguish: run takes a slug argument; import takes no slug and auto-selects best-fit.
• config/session-reports.md: filename convention generalized; resume/Continue Run logic scoped to Skill: job-tracker:run reports only.

v0.3.2

Added

• Network source layer: job-tracker:find network discovers opportunities from local referral/contact sources, and job-tracker:company checks local network sources before LinkedIn contact research.
• data/network/ as the canonical gitignored location for private contact/referral data, with committed README.md and .gitkeep.
• Source column for Raw Pipeline entries, including network for leads discovered through local contact sources.
• migrations/0.3.2.md — creates data/network/, adds the Raw Pipeline Source column, and adds network/referral paths to config/paths.md.

v0.3.1

Added

• Fit score bands and apply/skip verdict thresholds in search profiles (## Fit Score Bands in each profile); job-tracker:fit now outputs a structured Verdict + Recommendation based on profile thresholds (defaults: strong apply ≥45/60, apply with tailoring 35–44, low ROI <35).
• Migration mechanism for protected-zone evolution: CLI seeds missing files on init and update; job-tracker:health applies idempotent in-file migrations from migrations/<version>.md with confirmation; version markers config/.installed-version (CLI) and config/.migrated-version (health); works in plugin and workspace mode.
• migrations/0.3.1.md — first migration: adds ## Fit Score Bands to existing profiles, seeds candidate/application-answers.md in plugin mode, updates config/paths.md references.

Fixed

• Partial migration failure no longer bumps config/.migrated-version; health stops on first failed step and reports the exact failure so re-runs are safe.
• Seed steps skipped because source is missing are treated as failures (not satisfied); version marker is not bumped.
• bin/job-tracker.js: unused readFileSync import removed; seedIfMissingEntries now creates parent directories before copying.

v0.3.0

Added

• Plugin distribution as a Claude Code / Cowork plugin zip, built via npm run build:plugin and published as a GitHub release asset.
• Interactive scripts/install.js prompt (TTY) with non-TTY default to all for CI/npx use.
• Version tracking: CLI writes config/.installed-version on every init/update so agents can detect and propose upgrades.
• Pipeline Board section in job-tracker:status output.
• PDF import step in job-tracker:setup interactive onboarding (step 0).
• CSS customization guidance in job-tracker:setup (step 9).
• Version check step in job-tracker:setup that detects plugin vs workspace mode.

Changed

• Renamed all skill directories: job-run → run, job-setup → setup, etc. Commands are now /job-tracker:run, /job-tracker:setup, and so on, consistent between plugin and npx workspace modes.
• Renamed scripts/resume.css to scripts/cv.css; neutralized accent color and default styles so the file is a candidate-customizable starting point.
• style/cv-style.md neutralized — personal style rules removed.
• Plugin and npx workspace now share identical skill commands; no build-time namespace transforms needed.
• Stale skill directories (job-*) are now pruned from skills/, .claude/skills, and .codex/skills on update and install.
• update no longer overwrites scripts/cv.css (user-customizable); init copies it once.

Removed

• scripts/resume.css (replaced by scripts/cv.css).