main
4 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Corey Haines
|
a10be10a9f | merge main into development (vendor PRs that landed on main) before v2.2.0 release | ||
|
Corey Haines
|
f86637eace |
feat: add prospecting skill + truelist integration (#308)
* feat: add prospecting skill + truelist integration New skill: skills/prospecting/ - SKILL.md (251 lines, well under 500 limit): branch picker for SaaS / B2B / Local SMB, shared 5-phase framework (ICP -> discovery -> qualify -> score -> output), compliance guardrails, tool selection quick-picks, output formats - references/saas-prospecting.md: tech stack signals, funding/hiring triggers, SaaS-specific sources and qualification - references/b2b-prospecting.md: industry/firmographic signals, trigger events, decision-maker mapping, B2B-specific sources - references/local-prospecting.md: 4-tier website status classification, browser-assisted research workflow (generalized from the local-client- prospector pattern), proximity scoring - references/data-sources.md: deep dives on Apollo, Clay, ZoomInfo, Clearbit, Hunter, Snov, Truelist, LinkedIn Sales Nav, BuiltWith, Crunchbase, RB2B, with sequencing recommendations across the three branches - references/compliance.md: CAN-SPAM, GDPR, CASL, platform ToS (LinkedIn, Google Maps, Apollo/ZI/Clearbit), anti-patterns, audit checklist - evals/evals.json: 6 evals (2 SaaS, 2 B2B, 1 Local SMB, 1 deliverability) New integration: - tools/integrations/truelist.md: email deliverability validation (Deliverable / Risky / Undeliverable / Unknown classification) Registry + marketplace wiring: - tools/REGISTRY.md: truelist row + new Email Verification category section - .claude-plugin/marketplace.json: bumped to 2.1.0, prospecting added to plugin description - VERSIONS.md: prospecting 1.0.0 + 2.1.0 changelog entry - README.md: skill table re-synced, prospecting added to ASCII flow under Sales & GTM column All 41 skills pass validation. sync-skills.js is idempotent. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(prospecting): add GitHub stargazers/forks/watchers as discovery channel Net-new in this commit: - tools/clis/github-prospects.js: zero-dep Node CLI with commands stargazers / forks / watchers / user / rate-limit. Pagination via Link header, optional --enrich for full profile data, --with-email / --with-company / --with-blog filters, --format csv|json output, --dry-run preview. Uses GITHUB_TOKEN for 5000/hr rate limit (vs 60/hr unauthenticated). - tools/integrations/github.md: integration guide covering auth, rate limits, endpoints, workflows for SaaS prospecting, compliance notes (public API, not scraping), CLI reference. Skill updates: - skills/prospecting/SKILL.md: added GitHub to the tool selection quick picks and to the tool integrations table. - skills/prospecting/references/saas-prospecting.md: added GitHub to Tier 3 buying signals plus a dedicated "GitHub prospecting pattern (when audience is developers)" subsection with end-to-end workflow. - skills/prospecting/references/data-sources.md: added GitHub deep-dive section between RB2B and Free fallbacks. Registry: - tools/REGISTRY.md: github row in Tool Index, new Developer Intent / GitHub category section. All 41 skills still pass validation. sync-skills.js still no-op. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * refactor(prospecting): apply review suggestions CLI hardening + optimization: - github-prospects.js: encodeURIComponent on username path interpolation (defense in depth; GitHub usernames are restricted enough that this is safe in practice, but good hygiene). - github-prospects.js: refactored enrichUsers to filter inline and support --target N early termination. Previously, --with-email on a 1000-star repo would enrich all 1000 users before filtering down to the ~50 that match. Now you can pass --target 25 to stop as soon as 25 matches are found, saving API quota on restrictive filters. - github.md: documented the new --target flag. Reverse cross-references (so prospecting is discoverable from sibling skills): - cold-email: added prospecting as the natural upstream skill - customer-research: added "Translating customer research into an ICP for outbound" hand-off to prospecting - competitor-profiling: distinguished from prospecting ("this skill does deep research on specific accounts; prospecting builds the initial list") All 41 skills still pass validation. sync-skills.js still no-op. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(truelist): align integration doc with actual OpenAPI spec Source of truth: Truelist-Labs/truelist-openapi (OpenAPI 3.1). The earlier integration doc had inferred (and wrong) endpoint paths, request shapes, and status enum values. Corrected against the published spec: Base URL: https://api.truelist.io Endpoints (real): - POST /api/v1/verify_inline?email=... (sync single, email is query param) - POST /api/v1/verify (async bulk, body: {emails: [...]}) - GET /me (account info) Real email_state enum: - ok, email_invalid, risky, unknown, accept_all (not the inferred "Deliverable / Risky / Undeliverable / Unknown") Real email_sub_state enum: - email_ok, is_disposable, is_role, unknown_error, failed_smtp_check Also corrected: - Truelist has an official MCP server (Truelist-Labs/truelist-mcp) — was marked as MCP unavailable - Truelist has 7 official SDKs (Node, Python, Ruby, PHP, Go, Java, .NET) + framework integrations (Django, Laravel, Next.js, Rails, React, Svelte, Vue, WordPress) — was marked as SDK unavailable - Native integrations with Mailchimp, Klaviyo, HubSpot, Zapier, Make, n8n, Clay, Salesforce, ActiveCampaign, Brevo, ConvertKit, Drip, BigCommerce, Go High Level — was unlisted - Rate limits: 10 req/s per endpoint (was unspecified) Files updated: - tools/integrations/truelist.md: full rewrite against spec - tools/REGISTRY.md: MCP and SDK columns now show ✓ for truelist; classifier note in the Email Verification section reflects real enum values - skills/prospecting/evals/evals.json: eval #6 expected_output and assertions use real email_state values and mention the MCP server - skills/prospecting/references/data-sources.md: Truelist deep-dive uses real endpoint paths, real enum values, and lists the MCP/SDK ecosystem All 41 skills still pass validation. sync-skills.js still no-op. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * feat(prospecting): add Firecrawl + Browserbase for single-target site research Both tools are programmatic scrapers, but their use in prospecting is strictly bounded: extract content from individual public business sites (the prospect's own website URL), never from the platforms hosting them (Google Maps, LinkedIn, Yelp, Apollo, etc.). This matches the line drawn by the original local-client-prospector reference skill and our own compliance section. New integration docs: - tools/integrations/firecrawl.md: REST + MCP + SDKs (Node/Python/Go/Rust); scrape / map / crawl / extract / search endpoints; explicit "when NOT to use" section listing the prohibited platforms. - tools/integrations/browserbase.md: real Chromium via Playwright/Puppeteer or Stagehand (AI-friendly natural-language extraction); session recordings; useful when rendering or interaction is required. Prospecting skill updates: - SKILL.md: added Firecrawl + Browserbase to tool selection quick picks and tool integrations table. - references/data-sources.md: new "Firecrawl / Browserbase (single-target site research)" section between RB2B and Free fallbacks. Includes the compliance line inline so the framing isn't lost. - references/local-prospecting.md: optional "programmatic verification" paragraph in the browser research workflow — once you have a candidate's URL from manual Maps discovery, you can hit it programmatically. - references/compliance.md: anti-pattern #1 now explicitly clarifies that Firecrawl/Browserbase are fine for the prospect's own website but not for the platforms hosting prospects. Registry: - tools/REGISTRY.md: firecrawl + browserbase rows in Tool Index, new "Site Scraping (single-target only)" category section with the compliance framing in the agent recommendation. All 41 skills still pass validation. sync-skills.js still no-op. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com> |
||
|
erionjuniordeandrade-a11y
|
13fa45674c |
fix(ads): add Google RSA output spec with hard limits and ordering (#318)
The ads skill output was missing structural enforcement for Google RSA platform limits, causing AI agents to generate non-compliant ads: - variable headline/description counts (not exactly 15/4) - character lengths exceeded (>30 chars headlines, >90 chars descriptions) - missing negative keyword lists - missing ad group structure labels This commit adds a 'Google RSA Output Spec' section that: - Enforces 15 headlines (≤30 chars) / 4 descriptions (≤90 chars) per RSA - Requires explicit ad group structure, sitelinks (≥4), callouts (≥4) - Requires labeled negative keyword list (≥8 entries) - Specifies mandatory output ORDER (Negatives + sitelinks before RSAs) so they survive output truncation in long generations - Adds CFM (Brazilian medical regulation) forbidden-term list for pt-BR medical practice contexts - Adds a pre-response self-check checklist Validated on a marketing-eval harness across 3 medical sites: scores moved from 47/47/20 (all critical-fail) to 100/100/87 (all critical-pass) after applying the spec. Co-authored-by: Erion De Andrade <231011322+erionjuniordeandrade-a11y@users.noreply.github.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
Corey Haines
|
30f9b9a729 |
feat: v2.0 skill renames and CRO consolidation (#291)
* feat: v2.0 skill renames and CRO consolidation BREAKING CHANGE: Users must reinstall skills after this update. ## Skill Renames (16) - ab-test-setup → ab-testing - analytics-tracking → analytics - aso-audit → aso - competitor-alternatives → competitors - email-sequence → emails - free-tool-strategy → free-tools - launch-strategy → launch - onboarding-cro → onboarding - paywall-upgrade-cro → paywalls - popup-cro → popups - pricing-strategy → pricing - product-marketing-context → product-marketing - referral-program → referrals - schema-markup → schema - signup-flow-cro → signup - social-content → social ## Consolidations (1) - page-cro + form-cro → cro (form content in references/form.md) ## Why 2.0? - Shorter, cleaner skill names - Consistent naming (no -strategy, -setup, -cro suffixes) - All cross-references updated across 100+ files Total skills: 40 Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com> * fix(v2.0): update evals for renamed skills, fix validate script, clear warnings - Update 32 evals.json files to reference new skill names (page-cro → cro, product-marketing-context → product-marketing, etc.) — these were missed in the initial v2.0 rename pass since only SKILL.md and marketplace.json were updated. - Fix validate-skills.sh: replace GNU-only `head -n -1` with portable awk so frontmatter extraction works on macOS. - Move Copy Editing Checklist (56 lines) to references/checklist.md to bring copy-editing SKILL.md under the 500-line limit (508 → 457). - Add "see X" pointers to marketing-psychology description for skill discovery (cro, pricing, copywriting). - Update skill-request.yml issue template placeholder (page-cro → cro). All 40 skills now pass validation with zero warnings. * fix(v2.0): add evals for 8 missing skills, strip stale frontmatter from cro/form.md Adds 48 new eval cases (6 per skill) for skills that previously had no evals: aso, co-marketing, community-marketing, competitor-profiling, directory-submissions, image, lead-magnets, video. All 40 skills now have eval coverage (251 total cases). Strips leftover frontmatter from skills/cro/references/form.md — it was inherited from the old form-cro SKILL.md before consolidation. Reference files don't need frontmatter. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(v2.0): rename paid-ads → ads One more v2.0 simplification — drops the redundant 'paid-' qualifier. Updates the skill directory, SKILL.md frontmatter, evals.json, README skill table, the v2.0 rename table in VERSIONS.md (now 17 renames), and all cross-references in related skills (ad-creative, aso, competitor-profiling, customer-research, lead-magnets, marketing-ideas) plus the tools/integrations guides. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(v2.0): bump SKILL.md frontmatter version to 2.0.0 for all 40 skills VERSIONS.md was already updated to 2.0.0 but the metadata.version field inside each SKILL.md was still on 1.x. That mismatch would have caused the update-check flow to perpetually report 'update available' since it compares VERSIONS.md against local SKILL.md metadata versions. Caught by codex review (P1). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(v2.0): add legacy product-marketing-context.md filename fallback Before this fix, users upgrading from v1.x who had a `product-marketing-context.md` file would lose automatic context loading — every skill only checked the new `product-marketing.md` filename. Now all 40 skills also accept the legacy filename (in either `.agents/` or `.claude/`), and the README migration command covers both legacy and current filenames. Caught by codex review (P1 + P2). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(v2.0): re-sort README skills table alphabetically, fix ads box width The skills table had a few entries out of alphabetical order from the renames (co-marketing was after cold-email, ads was at the renamed position). Re-sorted alphabetically per sync-skills.js. Also padded the 'ads' cell in the ASCII flow diagram to keep the box width consistent. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fix(v2.0): document folder cleanup on upgrade, stop sync-skills from re-adding skills array README upgrade guide now includes: - A clear cleanup step for stale v1.x skill folders (renamed + consolidated) so users don't end up with both old and new folders side-by-side after upgrading - The full v1 to v2 rename map for reference - Existing product-marketing-context.md migration steps (preserved) sync-skills.js no longer (re-)introduces a `skills` array on marketplace.json -- Claude Code's plugin schema discovers skills via the `skills/` directory, and the explicit array was failing validation. The script now refreshes the description count and strips the stale array if present. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com> |