Brand voice built from evidence, a learnings ledger that drafting skills read, and two new loops. Voice and session-end patterns adapted with credit from Wispr Flow's MIT-licensed skills.
- product-marketing (2.1.0 → 2.2.0):
references/brand-voice.mdbuilds the voice profile from copy that performed and the founder's own writing: attributes with what each does and doesn't mean, tone by context (support, launch, errors, sales, social, billing), before/after examples, brand word lists, and per-person voice guides. The context doc gains an optional Paid Ads section (margin-based targets, hard floors, conversion truth, what the agent may change) and a Learnings table with promotion and retirement rules. - marketing-loops (1.5.0 → 1.6.0): an AI-answer accuracy audit loop (claims ledger with verdicts and severity, fix the source page first) and a weekly ads-check loop that ends each account in HOLD, FIX, TEST, or INVESTIGATE. Loop runs now close with a session-end step that logs 1–3 decisions and proposes learnings for approval; campaign postmortems feed the ledger. The catalog now has 58 loops.
- emails (2.1.3 → 2.1.4) and ad-creative (2.12.0 → 2.12.1): read the Learnings section before drafting.
Skill ZIP assets
Complete source catalog: 50 skills.
Claude web target: 50 skills with reviewed short descriptions.
Choose <name>-claude-web.zip for the documented single-folder web upload format.
Source ZIPs retain original metadata; they are not substitutes for omitted web ZIPs.
Live client upload, account limits, and invocation behavior have not been verified.
Web omissions (also recorded in claude-web-manifest.json):
- None.