Sent Agent Skills

Production-grade agent skills for SMS, WhatsApp, and RCS builders on Sent.

A Claude Code plugin that packages domain skills for the messy parts of running multi-channel business messaging on Sent — SMS 10DLC/TCR registration, WhatsApp template authoring and classification, RCS RBM agent onboarding, delivery-report analysis, multi-tenant Sender Profile architecture, the tenant template-builder UI, and Meta Embedded Signup.

---

What this is

Seven domain skills, matching slash commands, and the references they cite — packaged as an installable Claude Code plugin. Each skill encodes the decisions a senior messaging engineer makes in their head so an agent (Claude, Cursor, Gemini CLI, OpenCode, …) follows the same playbook consistently across all three channels Sent supports.

These skills are product-domain, not engineering-lifecycle. They compose cleanly with general engineering skills (e.g. addyosmani/agent-skills) — install both for full coverage.

---

Skills

Skill	Channel	Use it when…	Slash command
sender-profile-architect	all	Designing tenancy around Sender Profiles; webhook routing; rate limits across channels	/sender-plan
messaging-performance-analyzer	all	Investigating a delivery drop; finding the leak in a sales funnel across SMS/WhatsApp/RCS	/mdr-analyze
sms-10dlc-registration	SMS	Registering a brand + campaign with The Campaign Registry; debugging a TCR rejection	/sms-register
waba-template-author	WhatsApp	Drafting a template; classifying utility vs marketing; investigating a rejection	/waba-template
waba-embedded-signup	WhatsApp	Integrating Meta Embedded Signup; debugging a stuck signup	/waba-auth
rcs-agent-onboarding	RCS	Creating + verifying an RBM agent with Google; defining agent capabilities + fallback	/rcs-onboard
template-builder-ui	WhatsApp	Building the tenant-facing template editor; live preview; submission feedback	/template-ui

Plus /sent — a meta dispatcher that lists the skills and routes by intent.

---

Command ↔ skill mapping

Slash command	Skill
/mdr-analyze	messaging-performance-analyzer
/rcs-onboard	rcs-agent-onboarding
/sender-plan	sender-profile-architect
/sms-register	sms-10dlc-registration
/template-ui	template-builder-ui
/waba-auth	waba-embedded-signup
/waba-template	waba-template-author
/sent	sent (meta-dispatcher)

---

Quick Start

Claude Code — marketplace install (recommended)

/plugin marketplace add sentdm/agent-skills
/plugin install sent-skills@sent-skills

If marketplace clone fails due to SSH config, use the explicit HTTPS form:

/plugin marketplace add https://github.com/sentdm/agent-skills.git
/plugin install sent-skills@sent-skills

Claude Code — local / development install

git clone https://github.com/sentdm/agent-skills.git ~/dev/sent-agent-skills
# In Claude Code:
/plugin marketplace add file:///Users/$USER/dev/sent-agent-skills
/plugin install sent-skills@sent-skills

claude.ai (Pro / Max / Team / Enterprise)

# Zip an individual skill from the repo
cd skills && zip -r waba-template-author.zip waba-template-author/
# Then upload the .zip in claude.ai → Settings → Features → Skills

Repeat for each skill you want available.

Verify the install

In a fresh session:

1. Run /sent — the dispatcher should list every skill.
2. Ask "How do I classify a shipping-update template?" — the waba-template-author skill should auto-activate.
3. Ask "How do I register a 10DLC campaign?" — the sms-10dlc-registration skill should auto-activate.

---

How skills work

Skills are loaded by progressive disclosure (spec):

1. Discovery — at startup, only the skill name + description is in context (~100 tokens each).
2. Activation — when your request matches a skill's triggers, its SKILL.md body loads (under ~5k tokens).
3. Resources — referenced files (references/*.md) load only when the skill cites them.

This is why each skill's description aggressively includes trigger phrases — that's what the agent matches against.

---

Authoring a new skill

See CONTRIBUTING.md for the quality bar and docs/skill-anatomy.md for the format.

In short:

• One directory per skill under skills/<kebab-case-name>/
• One required file: SKILL.md with name + description frontmatter
• Description ≤ 1024 chars and must include both what + when
• Body ≤ 500 lines; push detail into references/
• Run bash scripts/validate-skills.sh before opening a PR

---

Repo layout

.claude-plugin/    Plugin + marketplace manifests
.claude/commands/  Slash command shims
skills/            Seven domain skills (one SKILL.md per directory)
  <name>/
    ├── SKILL.md
    ├── references/   ← per-skill reference docs
    └── scripts/      ← per-skill executable utilities (Python, stdlib only)
references/        Cross-cutting only (currently: sent-glossary.md)
docs/              Authoring + setup guides
scripts/           Repo tooling (skill validator)
.github/           Issue + PR templates, validate-skills workflow

Each skill directory is self-contained, so it can be zipped and uploaded to claude.ai as a standalone bundle.

---

Compatibility

Every skill follows the open Agent Skills format and works across any compatible agent — Claude Code, Cursor, Gemini CLI, OpenCode, and others. See AGENTS.md for cross-tool guidance.

---

License

MIT — see LICENSE.

Contributing

Pull requests welcome. Start with CONTRIBUTING.md. Bugs and proposals go in Issues.