Backed by: state/lib/knowledge.ts + state/lib/email.ts (draft only)

dormant-ping

Reads client-pulse output, filters to dormant_risk ≥ threshold_days, drafts a DM candidate per client via dispatch. Hard gate: writes to state/log/dormant-ping/<date>/<slug>.md and stops. Sending is a separate explicit skill.

Steps

1. Read latest client-pulse output (invoke if stale).
2. Filter by dormant_days ≥ threshold.
3. For each, dispatch gemini with {client_context, last_touch_excerpt}

→ DM draft.

1. Run every draft through voice.checkTone() before writing.
2. Write drafts; log chain row; exit.

Eval

Deterministic, three-part:

score("dormant-ping", run_id, {
  score:
    shape_ok && never_sent && all_drafts_pass_tone ? 1.0 :
    shape_ok && never_sent ? 0.5 :
    0.0,
  drafts_written: drafts.length,
  tone_violations_total: sum(drafts.map(d => d.violations.length)),
  primary_issue:
    !never_sent ? "auto-send-attempted" :
    !all_drafts_pass_tone ? "voice-gate-failed" :
    null,
});

The never_sent invariant - no send fn is imported, no HTTP POST happens - is verified by a lint check that greps this skill's run log for action: "deliver" rows (must be zero).

Gotchas

• voice.checkTone() gate is load-bearing. A "checking in" draft is

exactly the kind of copy that sprouts hype words ("quick hello, just wanted to supercharge our cadence") - the gate catches it.

• Threshold is per-client, not global. Read from

sources/clients/<slug>.md if present.

Rubric

criteria:
  - name: draft_files_written_correctly
    kind: deterministic
    check: "For each client meeting the threshold, a markdown file is written to 'state/log/dormant-ping/<date>/<slug>.md' with the DM draft."
  - name: never_auto_sent
    kind: deterministic
    check: "The run log for this skill ('state/log/dormant-ping/<date>/<run_id>.log') must contain zero instances of 'action: \"deliver\"'."
  - name: tone_check_passed_for_all
    kind: deterministic
    check: "The evaluation score for this skill must report 'tone_violations_total: 0'."
  - name: client_filters_correct
    kind: judge
    check: "The skill output appropriately filters clients based on 'dormant_risk ≥ threshold_days' and respects per-client thresholds from 'sources/clients/<slug>.md' if present."

dormant-ping - loader

Per-turn rules for the dormant-ping skill. Full reference: state/skills/dormant-ping/SKILL.md. Do not skip these.

Critical Rules

• NEVER auto-send. The skill writes drafts to state/log/dormant-ping/<date>/<slug>.md and stops. The never_sent invariant is verified by lint greppping for action: "deliver" rows (must be zero). Sending is a separate explicit skill.
• ALWAYS run every draft through voice.checkTone() before writing. A "checking in" draft is exactly the kind of copy that sprouts hype words ("quick hello, just wanted to supercharge our cadence") - the gate catches it.
• NEVER use scarcity framing ("idle hours", "leftover budget"). Value-first only. (feedback_no_scarcity_offers.md)
• NEVER open with personal stories ("I was…", "Last week I…"). Technical reports only. (feedback_no_personal_stories.md)

Commands

| ui dashboard | state/skills/dormant-ping/resources/ui.openui | |invoke: prose skill - read latest client-pulse output → filter dormant_days ≥ threshold → dispatch gemini per client → checkTone → write draft |verify: grep '"action":"deliver"' over state/log/chain.ndjson for this skill's run_id must return zero |eval log: state/log/evals.ndjson (auto eval - skill: "dormant-ping")

OpenUI Resource

• Skill-owned OpenUI Lang resource: state/skills/dormant-ping/resources/ui.openui. Read it before rendering or editing this skill's generated component surface.
• Treat this resource as a first-class artifact of the skill, not a generic chat response. Improve it when the skill's user-facing output needs to become richer.
• System resources compose OpenUI primitives and inherit SnappyChat tokens. Use ui_contract: branded in SKILL.md only for deliberate platform or client visuals.

Known Pitfalls

• Threshold is per-client, not global. Read from sources/clients/<slug>.md if present.
• Score is three-part: shape_ok && never_sent && all_drafts_pass_tone → 1.0; missing tone-pass → 0.5; missing never-sent (auto-send attempted) → 0.0

Self-Test

An agent reading this should correctly:

1. [ ] Refuse to import any send fn into this skill's run path
2. [ ] Apply voice.checkTone() to every draft before write
3. [ ] Read per-client threshold from sources/clients/<slug>.md, not a global default

Self-report

If this loader fell short, append a line:

echo "[$(date -u +%FT%TZ)] dormant-ping: <what was missing>" >> state/log/loader-feedback.log

<!-- kernel-ok: historical attribution for the PID-loader pattern. -->

Self-correcting loader (PID feedback)

This is the rule that makes loaders learn. Ported from snappy-kernel, where it demonstrably worked: the snappy-dom-cartographer skill mapped Skool admin UI once, every later ship-agent inherited the full map for free.

Before you finish, do two things:

1. Fix gaps (P - proportional). MANDATORY. If this loader didn't cover your case - if you had to read another file, run an undocumented command, work around a wrong selector, or discover a quirk - you MUST attempt an Edit to this AGENTS.md before you log.

• Default action: edit this file. Surgical. Add the missing line, table

row, or code example where it belongs. No restructuring.

• Only LOGGED is allowed when: the fix needs >10 lines, spans multiple

files, or requires a structural rewrite. In that case the state/regen/drain.sh queue picks it up asynchronously.

• "I didn't have time" / "it's minor" / "the next agent will figure it out"

are NOT valid reasons. Minor gaps compound - that's the whole point of this loop.

• Concrete example of a valid P-fix: discover agent-browser screenshot

takes path positionally not as --output <path>. One line in a table. Edit it in. Do not just log it.

• The goal: the next agent never has to leave the loader.

2. Log the result.

echo "[$(date -u +%FT%TZ)] dormant-ping: <what was missing or fixed> [FIXED|LOGGED]" >> state/log/loader-feedback.log

• FIXED = you patched this loader inline (P-fix).
• LOGGED = too large for inline; the PostToolUse enqueue + Stop-hook drain

will rewrite the loader from scratch on next session-end.

Do not skip this. Every agent run must leave the system better than it found it. The loader is the setpoint; you are the sensor; the gap is the error signal; closing the gap is the correction.