v0.10.1 - Clearbox-only Reddit source contract

The Reddit engine now has one explicit opportunity source: Clearbox. This patch removes the legacy parallel discovery path so the public architecture, runnable starter, and reusable agency skill all teach the same source contract.

Changed

• Clearbox-only source contract: pull.py imports a complete Clearbox export and requires the stable opportunity id, original lead/engage/competitor disposition, and exact Reddit permalink.
• Safe offline path: the bundled synthetic Clearbox export exercises the same contract without copying over a user's live export.
• Aligned public guidance: the engine guide, agency skill, AI-visibility playbook, proposal skill, examples, and GTM starter language now describe Clearbox as the authoritative opportunity source.
• Profile evidence terminology: the optional Playwright tier is described as a rendered-profile check, not as a Reddit collection path.

Removed

• Legacy discovery client and configuration: removed the external Reddit discovery client, its environment-key path, and the keyword/subreddit source files.

Verification

• Added importer tests for complete-export handling, truncation refusal, source-field validation, and exact permalink preservation.
• Added a release scan rule that rejects the retired source path from active Reddit engine surfaces.

v0.10.0 - The client pack you can see

[0.10.0] - 2026-08-10

The client pack you can see: a redesigned Clearbox dashboard, a real synthetic eleven-view demonstration, a guided public Notion brief, and an animated end-to-end README walkthrough backed by reproducible HyperFrames source.

Added

• Branded client-pack dashboard: the Sheet builder now renders a dense Clearbox dark/purple value surface with opportunity cards, offer and analysis coverage, signal-to-receipt workflow, evidence ladder, and guided next actions.
• Verified public demonstration: a view-only Acme Ops Sheet and guided Notion value brief generated from the repository's synthetic fixtures, with direct links from the README and gallery.
• End-to-end motion walkthrough: a short silent GIF, MP4, poster, frozen screenshots, shot plan, and checked HyperFrames composition showing Dashboard → Plan Setup → Operator Console → GEO receipts → Notion.
• Repository roadmap: product proof, broader gtm-coding-agent skill distribution, and the separately approval-gated launch campaign are explicit phases.
• Dashboard regression coverage: tests now assert the client-pack layout contract, its key values, evidence stages, twelve-column grid, and formatting request set.

Changed

• README visual story: the outdated five-tab screenshot and generic Notion guide no longer represent the current product. The README now leads with the animated eleven-view workflow, live synthetic demos, focused working-view captures, and the local/API build paths.
• Client-pack gallery: every new asset traces to the synthetic fixture build and a verified public surface; the earlier screenshots remain only as release-history references.
• Evidence language: the opening product explanation now states that search retrieval, observed AI appearance, exact citation, engagement, and business outcome require separate receipts.
• Coverage documentation: removed the stale “coming in v0.6.0” marker and linked the shipped enrichment-waterfall workflow.

v0.9.0 - Client value-pack automation

[0.9.0] - 2026-08-10

Client value-pack automation: any Clearbox user can turn the account API's classified opportunities and exact Reddit permalinks into the same guided Google Sheet and Notion delivery pattern used for agency clients.

Added

• API-to-client-pack builder: engine/build_client_pack.py pulls the account-scoped inbox or reads an export, preserves Clearbox dispositions and permalinks, and writes a normalized pack plus guided Notion-ready brief.
• Three interchangeable analysis adapters: sanitized Freckle JSON, Base Loop native JSON, and Clay CSV/JSON shapes normalize to the same output contract without replacing the original Clearbox disposition.
• Eleven-view Sheet contract: Dashboard, Plan Setup, Operator Console, Signals, Buyer Language, Content Topics, Competitor Sentiment, GEO Terms, Disclosure Audit, Research Workflow, and Action Legend, including decision and review dropdowns.
• Reusable client value-pack guide: build commands, Clay automation setup, stable Sheet/Notion ids, attribution receipts, truncation handling, client-safe language, and the release checklist.
• Pack validation tests: source disposition and permalink preservation, backend normalization, API read-only behavior, evidence fields, and Notion safety checks.

Changed

• reddit-agency skill: now generates one readable Notion value brief plus one working Sheet, with a short TLDR for sending and the detailed workflow kept in the guide.
• Agency access boundary: the public skill and builder are self-serve; the operated agency offering and multi-offer enablement require contacting partners@clearbox.to.
• Report automation boundary: API pulls, analysis merges, Sheet rebuilds, and Notion refreshes may be scheduled; Reddit publishing and opportunity completion remain human-authorized.

v0.8.1

[0.8.1] - 2026-08-10

Client-facing Plan Setup correction: the agency pack now explains the decision a client needs to make without exposing backend billing mechanics, admin controls, processing workspaces, or run identifiers.

Added

• Guided decision fields: Plan Setup requires dropdowns for the offer path, who pays, and readiness; operator queues use a Review Status dropdown.
• Collapsible Notion guides: the publisher supports toggle blocks so a client brief can explain every Sheet tab without becoming a wall of text.
• Client-copy release gate: the verification rules now reject provider-specific billing mechanics and internal workspace language in the public agency skill.

Changed

• reddit-agency 1.2.1: client copy now says an agency can add a separate client offer and the client can pay for it. Internal payment-provider behavior stays in admin context.
• Plan Setup language: one offer per client remains the operating rule, expressed through audience, queue, reporting, payer, and readiness instead of backend implementation details.

v0.8.0 - Plan and offer setup

[0.8.0] - 2026-08-10

Plan and offer setup: the agency skill now carries the actual multi-offering product behavior into every client command center instead of stopping at workflow and account guidance.

Added

• Plan Setup requirement: every agency command-center Sheet must recommend a path for the existing offer and explain when to add a separate offer for another client or service line.
• Offer and billing boundary: the operating guide now distinguishes the dashboard account, offer, plan, Reddit identity, operator, and billing owner.
• Verified multi-offering behavior: the guide records that each new offer gets its own Stripe customer and subscription, with billing independent across offers.

Changed

• reddit-agency 1.2.0: Step 0B and the delivery checklist now require a visible Plan Setup view and one isolated offer per client or genuinely separate service line.
• README stage table: added the plan and offer configuration stage to the public workflow map.
• Profile lookup secret loading: provider keys now come from the process environment only; the public engine no longer assumes private workstation databases or guesses secret-file paths.

v0.7.0

[0.7.0] - 2026-08-10

Profile lookup and orchestration: an evidence-bearing review gate that separates direct Reddit-profile disclosure, search candidates, no public evidence, and lookup errors, with per-tool integration guides for every enrichment backend.

Added

• Profile lookup skill: four-tier waterfall that checks the Reddit profile for exact disclosure evidence, then searches via Exa and DuckDuckGo for manual-review candidates. Every hit carries an evidence URL and excerpt; blocked checks remain lookup errors (skills/profile-lookup/, engine/lib/profile_lookup.py).
• Orchestrate enrichment playbook: the generic playbook describing the direct-disclosure gate, manual-review candidate path, retry path, and pluggable enrichment seam, with links to each tool integration guide (playbooks/orchestrate-enrichment.md).
• Base Loop integration guide: typed workflow with AI-powered stages and explicit fields for lookup status, review verdict, enrichment eligibility, and evidence receipts (examples/integrations/baseloop.md).
• Focused gate tests: direct Reddit-profile disclosure, search-only candidates, no-evidence versus error states, and enrichment eligibility (tests/test_profile_lookup.py).

Changed

• engine/unmask.py: only exact Reddit-profile company evidence can enter automatic enrichment. Search, thread-domain, and brand-handle matches are manual-review candidates. New --profile flag enables the full lookup waterfall; enrich_domain() remains the pluggable backend seam.
• README stage table: new rows for profile lookup and orchestration enrichment; Resolve who it is updated to reference the three-step gate.
• README playbooks list: orchestrate-enrichment.md added as the generic entry point above the tool-specific playbooks.

v0.4.0 — The skills drop

[0.4.0] - 2026-08-09

The skills drop: six new skills, three new engine scripts, a security showcase, and the attribution tracking pattern.

Added

• Competitor intel skill: share of voice, generated sentiment, and where the opening is — from classified Reddit ops. Includes FACTCHECK.md enforcing the classification-not-counting rule. Wraps engine/competitor.py + the new engine/sentiment.py.
• Sentiment skill: three-class (positive/neutral/negative) thread-level sentiment classification, LLM-powered with heuristic fallback. Output feeds competitor-intel and the daily digest.
• Dataviz skill: reference architecture for GTM dashboards with Recharts — the dark-theme palette, the component patterns, and three standalone .tsx examples (ChannelBars, ImpressionsLine, FunnelBars) plus a bootstrapping guide.
• Last 24 hours skill: the morning briefing — freshest buyer signals from the last day, ranked by intent and engagement.
• Reddit proposal skill: pitch materials for a prospect from their buyers' Reddit conversations. The reverse-uno applied to a single prospect.
• Slack digest skill: wraps engine/digest.py as a proper skill with setup, chaining, and the render-vs-post boundary.
• Three engine scripts: sentiment.py (thread-level sentiment from classified ops), last24.py (24h buyer signal feed), proposal.py (Reddit-sourced pitch brief).
• SECURITY.md: the protection model — siloed-by-design architecture (no cookies, no shared backend, local SQLite), the scan gate explained, read-only databases, the no-auto-send boundary, terminal examples, and the "using Reddit the right way" thesis.
• Attribution tracking playbook: the journey materialization pattern — identity spine, events table, first/last-touch channel attribution, and the "your data stays yours" architecture. Documents how the proof pipeline's signup-attribution tables are produced.

Changed

• README expanded with 8 new stage table rows (6 skills, attribution playbook, security doc) and rule #6 ("your data stays yours").
• ENGINE.md expanded with 3 new client-service modules.
• engine/digest.py now credits Lucas for the original daily digest pattern.

v0.3.0 — The flagship drop

[0.3.0] - 2026-08-07

The flagship drop: release hygiene, real branding, generated proof, transparency, and three new tracks (agency, students, partners).

Added

• Generated-not-typed proof pipeline: proof/generate_proof.py reads the tracking databases read-only and rewrites the stats block in proof/README.md between markers — with hard assertions on freshness (≤14 days), the exact views-claim language, and privacy. Includes the signup-attribution and inbound-engagement tables. No statistic in this repo is typed by hand anymore.
• Transparency: transparency/ — the honest channel ranking of what actually built the user base (what-actually-worked.md) and a full cold-email postmortem with the deliverability findings worth keeping (cold-email-postmortem.md).
• Offer-context onboarding as portable IP: playbooks/offer-context-onboarding.md — the interview pattern behind clearbox-onboard, generalized so any research agent can run it and any downstream system can consume the pack. clearbox-onboard bumped to 1.1.0 with the cross-links.
• Agency track: playbooks/win-an-agency-client.md — the reverse-uno method: lead with a readout of the client's buyers, not a pitch.
• Students: STUDENTS.md — a free semester of Clearbox Pro plus monthly group office hours for students who build in public, with the semester arc that produces a public track record.
• Partners: PARTNERS.md — referral and delivery tracks; terms by email.
• Two new skills: skills/geo-visibility/ (GEO term plan + retrieval-visibility score, wrapping engine/geo.py) and skills/longtail-content/ (buyer question → checked three-draft attribution pack, wrapping engine/content.py).
• Automation boundaries: playbooks/automation-boundaries.md — precisely what the machine may and may never do.
• Release hygiene: RELEASING.md, VERIFYING.md (the provenance rule, language boundaries, and the runnable scan gate every release passes), and a CI workflow that drafts a release when skills, playbooks, engine, proof, or transparency change.
• Brand assets: banner, social preview, and the official logo pack in assets/.

Changed

• README rebuilt as the flagship map: the full stage table (onboarding featured as the differentiator), transparency section, and consolidated related links.
• The public multi-account guide on Notion and this repo now link to each other bidirectionally.

v0.2.0: Multi-account operations

ClearboxGTM now includes the reusable agency operating system for running multiple client Reddit accounts safely and measuring the work honestly.

Highlights:

• Client-owned identities, named operators, separate workspaces, and coordination controls
• A specific decision model for one owner with a SaaS company and a related services business
• A dated official-source and public-practitioner evidence ledger separating fact, observation, fiction, and unproven claims
• A five-level scorecard for Reddit artifact health, search discovery, observed AI answer receipts, retrieval visibility, and business outcomes
• A reusable AI visibility receipt CSV
• Engine terminology corrected so Exa retrieval is not misreported as an AI citation

Public client guide: https://fierce-camelotia-1fa.notion.site/Clearbox-Running-Multiple-Reddit-Accounts-for-Clients-3b51fb92bcd78187a212de323c577399

Full changes: https://github.com/shawnla90/ClearboxGTM/blob/v0.2.0/CHANGELOG.md