Promo Cycle Monitor

A pipeline-aware promotion cycle monitoring skill that doesn't just count open assignments — it classifies their risk state and surfaces the ones that will become escalations if no one acts today.

Most promotion cycle tooling is reactive. You find out a panel disappeared after the deadline passed. You find out a panelist never responded after the candidate has been waiting three weeks. You find out about a {{NOMINATION_PLATFORM}} access issue when the panelist emails you in frustration.

This skill is architecturally different. It was designed using Adaptive Narrative Control Theory (ANCT) — each phase uses the control mode matched to its actual uncertainty level:

• Pipeline pull → DELEGATE (low entropy, just enumerate)
• Risk classification → NARRATE (medium entropy, requires judgment about what "stalled" means)
• Exception triage → NARRATE (E3, policy-governed decisions about edge cases)
• Follow-up drafting → GENERATE → NARRATE (high entropy, situation-specific messaging)
• Status snapshot → NARRATE → DELEGATE (compress findings into an actionable dashboard)

The insight: The most dangerous assignment in a promotion cycle is not the one with a red flag — it's the one with no flag at all because it silently expired. The skill is designed to find those first.

Core Philosophy

• Silence is a signal — a panel with no activity after 5 days is not fine, it is at risk
• Overdue is not one category — some overdue panels need a nudge; some need escalation; some are already gone
• Volume is the enemy of oversight — 110+ assignments means no one catches the dropped ones manually
• Classification before action — the wrong follow-up (remind vs. escalate vs. reassign) makes things worse
• The pipeline is the product — the output is not a report, it is a current-state view of what needs a decision today

ANCT Architecture

This skill was designed with the adaptive-workflow-architect meta-skill. Here is its entropy profile and mode map:

Phase:      1          2          3          4          5            6
Entropy:    E1         E2         E3         E4         E3           E1→E2
Mode:       DELEGATE   DELEGATE   NARRATE    GEN→NAR    NARRATE      NAR→DEL

            pull       classify   surface    draft      triage       produce
            pipeline   by status  risks      follow-ups exceptions   snapshot
            from                  and stale             (access,     (counts,
            inbox/                items                 post-        flags,
            {{NOMINATION_PLATFORM}}                                     deadline)    decisions)

Mode Transition Diagram

┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────────────────┐
│ DELEGATE  │→│ DELEGATE  │→│ NARRATE   │→│ GENERATE → NARRATE    │
│ Pull all  │  │ Classify  │  │ Surface   │  │ Draft follow-ups      │
│ open      │  │ each      │  │ at-risk,  │  │ per panelist and      │
│ panels    │  │ assignment│  │ overdue,  │  │ situation type        │
│ from      │  │ on-track/ │  │ silently  │  │                       │
│ inbox +   │  │ at-risk/  │  │ dropped   │  │ ⚡ COMPRESSION        │
│ {{NOMINATION_PLATFORM}}   │  │ stalled/  │  │ items     │  │ CHECKPOINT: one       │
│           │  │ dropped   │  │           │  │ action per assignment │
│ E1        │  │ E2        │  │ E3        │  │ required before E5    │
└──────────┘  └──────────┘  └──────────┘  └──────────────────────┘
                                                       │
                                                       ▼
                              ┌──────────┐  ┌──────────────────────┐
                              │ NAR→DEL   │←│ NARRATE               │
                              │ Produce   │  │ Triage exceptions:    │
                              │ status    │  │ post-deadline noms,   │
                              │ snapshot  │  │ access issues,        │
                              │ dashboard │  │ policy edge cases     │
                              │           │  │                       │
                              │ E1→E2     │  │ E3                    │
                              └──────────┘  └──────────────────────┘

Intake: Customize This Skill

{{CYCLE_NAME}}

{{PANEL_TYPE}}

{{LOOKBACK_DAYS}}

{{STALE_THRESHOLD_DAYS}}

{{INBOX_SOURCE}}

{{COMPENDIUM_PATH}}

{{NOMINATION_PLATFORM}}

Phase 1: PULL PIPELINE

Control Mode: DELEGATE | Entropy: E1 (Deterministic)

This is pure enumeration. No judgment. Pull everything, classify nothing yet.

Actions

1. Pull from

   {{INBOX_SOURCE}}

   all emails matching

   {{CYCLE_NAME}}

   within

   {{LOOKBACK_DAYS}}

   :
   • Panel assignment notifications
   • Panelist response emails (accepted, declined, completed)
   • Deadline reminder messages
   • {{NOMINATION_PLATFORM}} access request threads
   • Exception request emails (post-deadline nominations, panel swaps)
   • Escalation threads

2. Pull from {{NOMINATION_PLATFORM}} (if accessible) or

   {{COMPENDIUM_PATH}}

   (if set):
   • Open panel assignments by candidate
   • Assignment status per panelist
   • Deadline dates per assignment
   • Completion timestamps where present

3. Deduplicate and merge inbox signals with {{NOMINATION_PLATFORM}} records. One row per assignment.

4. Enumerate total counts by raw status bucket:

   • Pending — assigned, no response yet
   • Accepted — panelist confirmed, evaluation not yet submitted
   • Submitted — evaluation complete
   • Overdue — past deadline, not submitted
   • Expired — deadline passed with no panelist activity (no accept, no decline, no submission)
   • Declined — panelist declined or was removed

Exit Condition

All assignments for

{{CYCLE_NAME}}

Output

**Pipeline Pull — {{CYCLE_NAME}} ({{PANEL_TYPE}})**
**Source:** [inbox / {{NOMINATION_PLATFORM}} / compendium]
**Pulled:** [N] total assignments

| Status    | Count |
|-----------|-------|
| Pending   | N     |
| Accepted  | N     |
| Submitted | N     |
| Overdue   | N     |
| Expired   | N     |
| Declined  | N     |

**{{NOMINATION_PLATFORM}} access:** [available / unavailable — continuing from inbox]

Phase 2: CLASSIFY STATUS

Control Mode: DELEGATE | Entropy: E2 (Procedural)

Same classification logic applied to every assignment. No judgment yet — that is Phase 3's job. Apply the rules consistently.

Classification Rules

{{STALE_THRESHOLD_DAYS}}

{{STALE_THRESHOLD_DAYS}}

Escalation → NARRATE

If an assignment's state does not fit cleanly into the above table (ambiguous email signals, conflicting {{NOMINATION_PLATFORM}} vs. inbox data, partial completion records):

• Flag the assignment as ambiguous
• Carry it into Phase 3 for manual classification judgment

Exit Condition

Every assignment has exactly one classification. Ambiguous assignments are flagged and carried forward. No assignment remains unclassified.

Output

**Classification Results — {{CYCLE_NAME}}**

| Classification             | Count |
|----------------------------|-------|
| on-track                   | N     |
| at-risk                    | N     |
| stalled                    | N     |
| silently-dropped           | N     |
| completed                  | N     |
| declined-needs-reassignment| N     |
| ambiguous (flagged)        | N     |

Phase 3: SURFACE RISKS

Control Mode: NARRATE | Entropy: E3 (Analytical)

This is where judgment begins. Not every at-risk assignment is the same risk. Not every overdue item needs the same action. Prioritize by impact and urgency.

Actions

1. From the Phase 2 classifications, pull all assignments that are:

   • silently-dropped (highest priority — these are invisible without this step)
   • stalled (deadline passed, panelist accepted but did not submit)
   • at-risk (approaching deadline or unresponsive panelist)
   • ambiguous (unresolved from Phase 2)
   • declined-needs-reassignment (panel is uncovered)

2. For each flagged assignment, assess:

   • How long has it been since any activity? (days since last signal)
   • Is the candidate's promotion decision blocked by this panel?
   • Has this panelist gone silent on other assignments in the same cycle?
   • Is this a pattern (same panelist, same manager's nominees) or isolated?
   • Is there an upcoming deadline cascade (multiple panels expiring same week)?

3. Prioritization (this is the NARRATE work):

   • Priority 1: silently-dropped + candidate decision imminent
   • Priority 2: stalled + panelist unresponsive > 7 days
   • Priority 3: at-risk + deadline within 48 hours
   • Priority 4: declined-needs-reassignment (panel uncovered)
   • Priority 5: ambiguous (needs human resolution)

4. Surface the top items. Do not list all 110+ assignments — surface the ones that require a decision or action today.

Escalation → GENERATE (if no risks surface)

If Phase 2 classifications show all on-track or completed and no risks surface:

"No at-risk or dropped assignments detected. Given the cycle volume and typical patterns, what assignments might be misclassified as on-track? What signals in the inbox data are ambiguous?"

Generate a skeptical secondary pass rather than accepting a clean state at face value.

Exit Condition

Prioritized list of assignments requiring action, with reasoning for priority. Or: generated skeptical review of the clean-state result.

Output

**At-Risk Assignments — {{CYCLE_NAME}}**

Priority 1 — Silently Dropped (candidate decision imminent):
- [Candidate] / [Panelist] — last activity: [N days ago] — deadline: [date] — [note]

Priority 2 — Stalled (accepted, not submitted, unresponsive):
- [Candidate] / [Panelist] — accepted [date] — [N] days since last contact

Priority 3 — At Risk (deadline within 48h):
- [Candidate] / [Panelist] — deadline: [date/time] — status: [accepted / no response]

Priority 4 — Uncovered Panels (needs reassignment):
- [Candidate] — panelist [name] declined [date] — no replacement assigned

Priority 5 — Ambiguous (needs human judgment):
- [Candidate] / [Panelist] — conflict: [describe inbox vs. {{NOMINATION_PLATFORM}} discrepancy]

Phase 4: GENERATE FOLLOW-UPS

Control Mode: GENERATE → NARRATE | Entropy: E4 (Creative)

This is the highest-entropy phase. The same follow-up message sent to a panelist who forgot and a panelist who is blocking on a {{NOMINATION_PLATFORM}} access issue will get different responses — the wrong one wastes time or creates friction.

GENERATE Step (expand)

For each Priority 1-3 assignment from Phase 3, generate 2 follow-up drafts based on the most likely situation:

Draft A (default): Polite reminder — assumes the panelist forgot or deprioritized
Draft B (escalation): Situation-aware — acknowledges a specific barrier
                      (access issue, unclear instructions, conflict)

Signals to weigh per panelist:

• No activity at all (never responded to assignment) → Draft A first, then B if no response
• Accepted but no submission + long gap → ask if there is a blocker (Draft B)
• Previous {{NOMINATION_PLATFORM}} access thread in inbox → Draft B immediately (access is the known barrier)
• Pattern across multiple candidates → flag for manager escalation, not another nudge
• Same panelist, multiple silently-dropped → escalate upward, do not re-remind

COMPRESSION CHECKPOINT

Before proceeding to Phase 5, each at-risk assignment must have exactly one recommended action:

send Draft A / send Draft B / escalate to manager / reassign panel / flag for human decision

No assignment leaves Phase 4 without a recommended action.

NARRATE Step (compress)

Select the most appropriate draft per assignment. Annotate why. Surface any assignments where both drafts are inadequate and human judgment is required.

Exit Condition

One recommended action and one draft message per at-risk assignment. Assignments requiring escalation beyond a message are separated into an escalation list.

Output

**Follow-Up Drafts — {{CYCLE_NAME}}**

---
**[Candidate] / [Panelist]** | Action: Send Draft A
> Subject: Promotion Panel — {{CYCLE_NAME}} — [Candidate Name]
>
> Hi [Panelist],
>
> Just following up on the panel assignment for [Candidate] as part of
> {{CYCLE_NAME}}. The evaluation window [closes / closed] on [date].
>
> If you have any questions or need access to {{NOMINATION_PLATFORM}}, reply here and
> I can help get that sorted quickly.
>
> Thanks,
> [Your name]

---
**[Candidate] / [Panelist]** | Action: Send Draft B (known access issue)
> Subject: Re: {{NOMINATION_PLATFORM}} Access — {{CYCLE_NAME}} Panel
>
> Hi [Panelist],
>
> I saw the earlier thread about {{NOMINATION_PLATFORM}} access. If that's still
> blocking you on the [Candidate] panel, I want to make sure we get
> it resolved before [deadline]. Can you confirm whether access is
> working or if I should re-open a ticket?
>
> [Your name]

---
**[Candidate] / [Panelist]** | Action: ESCALATE — pattern detected (3 assignments, no activity)
> Recommend: Reach out to [Panelist]'s manager directly.
> Do not send another reminder — prior reminders have not been effective.

Phase 5: EXCEPTION TRIAGE

Control Mode: NARRATE | Entropy: E3 (Analytical)

Exceptions are the cases where the standard process does not apply and a policy-governed decision is required. These are not failures — they are judgment calls that need to be surfaced explicitly so a human can decide.

Exception Types to Triage

Actions

1. Identify all exception threads from Phase 1 inbox pull.
2. For each exception, classify the type using the table above.
3. Determine whether the exception has already been resolved (look for a reply confirming resolution).
4. Surface unresolved exceptions with the decision that is required.
5. Do not make the policy decision — surface it for human action.

Exit Condition

All exceptions identified, classified, and resolution-status assessed. Unresolved exceptions surfaced with the specific decision required.

Output

**Exception Triage — {{CYCLE_NAME}}**

Unresolved Exceptions Requiring a Decision:

1. [Type: Post-deadline nomination]
   Candidate: [Name] | Nominated by: [Manager] | Date: [N days after deadline]
   Decision required: Accept with justification / Decline

2. [Type: {{NOMINATION_PLATFORM}} access failure]
   Panelist: [Name] | Candidate: [Name] | IT ticket: [open / not yet filed]
   Decision required: Extend deadline / Accept off-system / Reassign

3. [Type: Incomplete panel]
   Candidate: [Name] | Submitted: 1 of 2 panelists | Missing: [Panelist]
   Decision required: Proceed with single submission / Hold for second

Resolved Exceptions (no action needed):
- [Candidate] — panel swap approved, replacement [Name] assigned [date]

Phase 6: PRODUCE STATUS SNAPSHOT

Control Mode: NARRATE → DELEGATE | Entropy: E1 → E2

Synthesize all phases into a single cycle dashboard. This is the artifact that gets shared, acted on, or filed. Format is consistent — this is the output that builds trust because it looks the same every time you run it.

Dashboard Structure

## Promotion Cycle Status Snapshot
**Cycle:** {{CYCLE_NAME}} | **Panel Type:** {{PANEL_TYPE}}
**Run date:** [date] | **Lookback:** {{LOOKBACK_DAYS}} days

---

### Pipeline Summary

| Status             | Count |
|--------------------|-------|
| Completed          | N     |
| On-Track           | N     |
| At-Risk            | N     |
| Stalled            | N     |
| Silently Dropped   | N     |
| Needs Reassignment | N     |
| Ambiguous          | N     |
| **Total**          | **N** |

---

### Red Flags — Action Required Today

1. [Assignment] — [Classification] — [Recommended action] — [Deadline]
2. [Assignment] — [Classification] — [Recommended action] — [Deadline]
3. [Assignment] — silently dropped — escalate immediately

---

### Pending Decisions

| Exception | Type | Decision Needed |
|-----------|------|-----------------|
| [Candidate] | Post-deadline nomination | Accept / Decline |
| [Panelist] | {{NOMINATION_PLATFORM}} access | Extend / Reassign |

---

### Follow-Ups Staged (ready to send)

- [N] Draft A reminders
- [N] Draft B situation-aware messages
- [N] Escalation recommendations (do not send another reminder)

---

### Cycle Health Indicator

[GREEN / YELLOW / RED]

GREEN:  >80% completed or on-track, <3 silently dropped, no unresolved exceptions
YELLOW: 15-30% at-risk or stalled, or 3-7 silently dropped, or 1-2 unresolved exceptions
RED:    >30% at-risk/stalled, or >7 silently dropped, or exception cascade in progress

Anti-Patterns

PROMO CYCLE MONITOR ANTI-PATTERNS (from ANCT failure mode analysis):

X "Remind again" as default for all overdue
  Every overdue assignment gets the same nudge email.
  No distinction between a panelist who forgot and one who is blocked on {{NOMINATION_PLATFORM}}.
  → ANCT diagnosis: flat DELEGATE applied to E4 phase; skipped GENERATE step
  → Correct: classify the barrier first, then select Draft A or Draft B

X Marking silently-dropped panels as complete
  {{NOMINATION_PLATFORM}} shows the assignment as closed. No submission record exists.
  Assumes the system is correct and the panel was completed.
  → ANCT diagnosis: DELEGATE applied where NARRATE was required;
    absence of completion record was not treated as a signal
  → Correct: flag as silently-dropped; require positive confirmation of completion

X Clean dashboard accepted at face value
  All assignments show on-track. No further investigation.
  In a 110+ assignment cycle, some are always misclassified.
  → ANCT diagnosis: premature compression; skipped skeptical GENERATE pass
  → Correct: if no risks surface, run a secondary skeptical scan (Phase 3 escalation)

X Treating all exceptions as standard delays
  A post-deadline nomination is processed the same as a late submission.
  → ANCT diagnosis: exception triage skipped; E3 judgment replaced with E1 rule
  → Correct: surface exceptions separately; make the policy decision explicit

X Listing all 110+ assignments in the output
  Volume obscures the actual risks.
  → ANCT diagnosis: Phase 3 NARRATE filter not applied; DELEGATE logic returned all rows
  → Correct: surface only items requiring action; snapshot shows counts, not rows

Example Applications

{{STALE_THRESHOLD_DAYS}}

Platform Notes

{{COMPENDIUM_PATH}}

/skill promo-cycle-monitor

Cross-References

This skill is designed to work with related skills in the library:

{{COMPENDIUM_PATH}}

How to Use Together

1. Run

   inbox-intelligence

   on the promotion cycle inbox folder before running this skill
2. The compendium indexes all panel assignment threads, {{NOMINATION_PLATFORM}} notifications, and panelist replies
3. Set

   {{COMPENDIUM_PATH}}

   when invoking

   promo-cycle-monitor

4. Phase 1 queries the compendium instead of raw inbox — faster and cross-referenced
5. Phase 4 (follow-up drafts) references prior interaction history from the compendium to avoid re-sending messages that were already sent

Compliance

• PII Risk: Medium. Panelist names, candidate names, evaluation status, and communication history appear in all phase outputs. Store the status snapshot in a private compendium only. Do not paste output into shared channels or public interfaces.
• Retention: Cycle snapshots should be archived per your organization's talent data retention policy. Do not retain candidate or panelist data after the cycle closes unless required for audit.
• Model Recommendation: Sonnet for full runs (Phase 4 follow-up generation and Phase 5 exception triage benefit from strong reasoning). Haiku is acceptable for Phase 6 snapshot-only runs on a clean pipeline.
• Human Oversight: The skill surfaces risks and drafts messages — it does not send them. Every follow-up and every exception decision requires human review before action. The recommended action per assignment is a recommendation, not an execution.

Design Credit

This skill's architecture was designed using the adaptive-workflow-architect meta-skill, applying Adaptive Narrative Control Theory (ANCT) to map each phase to its optimal control mode based on entropy level.