Skip to content
Agentmart
Submit a tool

trust-ladder-workflow

Per-company trust state procedure (learning / trial / operating).

Skill by petr-royce

IndexedMIT★ 106
Write a review

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/elophanto-trust-ladder-workflow

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Per-company trust state procedure (learning / trial / operating). When a live outreach tool refuses, names the canonical draft replacement.

Use cases: sales, productivity

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/trust-ladder-workflow

What's inside

Skill name
trust-ladder-workflow
Description
Per-company trust state procedure (learning / trial / operating). When a live outreach tool refuses, names the canonical draft replacement.
License
MIT
Metadata
phase: 9 · author: petr-royce · version: 1.1.0
Instructions
61 lines in SKILL.md · view on GitHub

Files (1)

  • SKILL.md

SKILL.md (first lines)

# Trust ladder workflow — DO NOT spam

Every ABE company has a 3-state trust ladder gating live outbound communication:

| State | Live outreach | Per-call approval | Default |
|---|---|---|---|
| `learning` | **REFUSED** — draft only | n/a | new companies |
| `trial` | allowed | MODERATE | promoted after voice approval |
| `operating` | allowed | within budget | mature |

## Triggers

- email_send / email_reply / prospect_outreach / twitter_post
- send email / send outreach / post on X
- can you message / can you reach out
- trust state / trust ladder / promote trust

## Procedure

1. Check trust state for active company (`company_report` headline or `[COMPANY] trust=` in snapshot).
2. **If `learning`**: live tool refuses. Call the draft equivalent:
   - `email_send`/`email_reply` → `email_draft(to, subject, body)`
   - `prospect_outreach` → `outreach_draft(prospect_id, body, channel)`
   - `twitter_post` → `post_draft(content)`
   
   Drafts land at `companies/<slug>/drafts/<kind>/pending/`. Voice contract lints body before persist (see [[voice-extraction-workflow]] if lint fails). Tell operator what you drafted.

3. **If `trial`**: live tools work but each gates MODERATE. Be deliberate — operator just promoted you. Produce ONE quality send, not a batch.
4. **If `operating`**: full autonomy within budget. Voice contract still lints.

## When a live tool refuses

Error message names the replacement and operator action. Example:
```
email_send blocked: company 'acme' in trust state 'learning' — use email_draft.
Operator command: elophanto company trust acme trial.
```
- Don't retry the live tool (gate is operator-decided, not transient).
- Call the named draft tool.
- Tell operator what you drafted + remind them of the promotion command.

## Promotion (operator only)

- **Never call `company_trust_set` yourself.** Operator decides via CLI or via the tool with explicit MODERATE approval.
- Surface readiness signals: voice.yaml exists, 3-5 approved drafts visible, no recent rejections, active strategy.
- After operator promotes to `operating`: live sends become autonomous. Demote with `elophanto company trust <slug> trial` if it goes sideways.

## Hard rules

- ❌ Never `company_trust_set` to promote yourself.
- ❌ Never re-call a live tool after trust-gate refusal.
- ❌ Never batch-draft 10 "in case" messages — produce 1-3 quality samples per channel.
- ❌ Never silently swap channels when one refuses.

## Verify

- [ ] Trust state checked before any live outbound call
- [ ] On gate refusal: draft tool called, not retried
- [ ] On promotion request: I did NOT call `company_trust_set`

From elophanto/EloPhanto, MIT.

More from elophanto/EloPhanto

Signals

Stars
106
Forks
9
Open issues
2
Last commit (this folder)
May 26, 2026 (5 months ago)
Last release
Oct 1, 2026 (27 days ago)

From GitHub (fetched Oct 24, 2026).

Details

Version
1.1.0
License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from petr-royce

Cite this page

trust-ladder-workflow skill. Agent Mart. https://agentmart.dev/skills/elophanto-trust-ladder-workflow (last updated 2026-10-24)