Marketing Skills v2.4.0

A minor release adding one new skill (public-relations, originally shipped as pr — renamed in v2.4.1) for earned media work, plus a substantial new social listening workflow added to the existing social skill.

Total skills: 44

New skill: public-relations

Earned media work for software companies — pitching journalists, newsjacking trending stories, responding to press requests (HARO/Connectively, Qwoted, Featured, Help A B2B Writer), and building the owned-media foundation (press page + media kit).

The 4 PR modes

The skill is built around four modes most teams skip three of:

1. Reactive (newsjacking) — Inject your POV into trending news. 4–12 hour window for breaking tech stories. Score → angle → pitch loop with a newsworthiness rubric. Auto-disqualifies tragedy-jacking and forced fits.
2. Proactive (journalist pitching) — Build a media list of 20–40 named journalists, score for fit, pitch original stories using one of six angle templates.
3. Inbound (press platforms) — Daily triage workflow for HARO/Qwoted/Featured/Help A B2B Writer. Includes a quoteable response template and a realistic ROI check (5 hr/week → 3–10 placements; 1 hr/week → 0–2 — skip if you can't sustain).
4. Owned (press page + media kit) — Cheapest PR investment with highest ROI. Press page checklist + media kit assets list. One afternoon, then forget it.

Six pitch templates by angle

Each template is under 150 words and has a specific structure:

• Data story — "[Specific stat] — [implication]" subject + N-analyzed/surveyed setup + 3 stats + the most interesting pattern + raw data offer
• Exclusive launch — "Exclusive: [specific milestone]" subject + embargo offer
• Op-ed pitch — "Op-ed pitch: [provocative thesis]" subject + thesis + 3 supporting points + credential + news hook
• Customer story — "Customer story for [their beat] — [specific outcome]" subject + named customer willing to talk on the record
• Trend connector — "Trend forming in [space] — three signals" subject + three connected signals + the pattern
• Newsjack response — "Re: [their article headline] — quick data point" subject + one-sentence stat + methodology + quotable line + phone for the day

Each pitch is banned from "revolutionary," "disruptive," "game-changing," "synergy," and a longer banned-vocabulary list. The skill won't let you ship a pitch using them.

Reference framework (4 docs)

• newsjacking.md — detect → score → angle → pitch loop, 7-dimension newsworthiness scoring rubric, 7 story-angle templates, curl recipes for Google News RSS / HN Algolia / Reddit, failure modes ("tragedy-jacking," "forced fit," "empty take")
• journalist-pitching.md — media list construction, 4-dimension journalist fit scoring rubric, the 6 pitch templates, subject-line patterns that work vs. patterns that get deleted, voice and structure rules, embargo and exclusive etiquette, follow-up cadence, tooling (Hunter.io, RocketReach, dev-browser for journalist research)
• press-platforms.md — daily triage workflow for HARO/Connectively, Qwoted, Featured, Help A B2B Writer, SourceBottle, Terkel, JournoRequests; response template; what makes a response get selected vs. killed; ROI reality check by effort tier; setup checklist
• media-outlets.md — tiered list of tech press, SaaS, AI, devtools, business, plus newsletters and podcasts — explicitly scoped to journalist-driven outlets, with startup directories deferred to the directory-submissions skill

Scope is enforced via cross-references

• public-relations handles earned media
• directory-submissions handles Product Hunt / BetaList / SaaS directories
• launch handles the broader launch moment
• social handles social-media engagement (with the new listening workflow below)

Updated skill: social (2.0.0 → 2.1.0)

Added a listening workflow for daily engagement triage — surfacing the top 10 posts to comment on each morning rather than scrolling feeds.

references/listening.md

Daily loop for engagement: pull from defined sources → filter → score → draft → post → log.

• 5-dimension scoring rubric — ICP fit (2x), intent signal (2x), reach potential (1x), comment opportunity (2x), recency (1x)
• Comment quality tiers — Tier 1 relationship builder (target accounts, ICP, high intent), Tier 2 visibility play, Tier 3 light touch
• Curl recipes — Reddit JSON API, Hacker News (Algolia), Bluesky public API, RSS for blogs/YouTube channels — all free, no auth required
• Browser-driven workflows — dev-browser MCP with persistent session for LinkedIn and X (no good public API exists), with the specific URLs to feed (linkedin.com/in/HANDLE/recent-activity/all/, x.com/i/lists/LIST_ID, etc.)
• Per-platform notes — LinkedIn algorithm-friendly comment patterns, Twitter/X timing windows, Reddit subreddit etiquette, Hacker News commenting bar, Bluesky engagement-to-follower ratio

references/listening-sources-template.md

Copyable starter for .agents/listening-sources.md covering: brand/category context, ICP definition (for scoring), target accounts split by platform (LinkedIn/X marked browser-driven, others scriptable), intent keywords with concrete examples, subreddits, saved-search URLs, and a do-not-engage list.

SKILL.md Engagement Strategy section gets a pointer to the listening reference.

Notes on the release

The initial v2.4.0 release shipped with broken plugin metadata (marketplace.json and plugin.json were stuck at v2.3.0, so claude plugin update wouldn't surface this release). The release notes here describe the corrected state after the metadata bump and a follow-up hot-fix that trimmed the social skill description back under the 1024-char Agent Skills limit.

Subsequent renames and corrections were folded into v2.4.1 (the pr → public-relations rename) — see that release for details.