receipt-stream

Receipt-Stream — cross-cycle ε feed (per-cycle iteration artifact + cross-cycle aggregator)

Core principle

The receipt-stream is the protocol's own learning signal across cycles. Each cycle that surfaces a protocol gap (protocol_gap_count > 0) emits a per-cycle iteration artifact (cdd-iteration.md) with a fixed per-finding shape; γ then appends one row to the cross-cycle aggregator (.cdd/iterations/INDEX.md). ε reads the aggregator to surface patterns across cycles; the per-cycle artifact carries the structured findings the aggregator points at. Cycles with protocol_gap_count == 0 may write a courtesy empty-findings stub (per cycle/401 convention) but are not required to.

The per-cycle artifact is role-local: γ authors it as part of the cycle's close-out (with ε review where ε is a distinct actor; ε=δ collapsed in the current cnos cdd operating point per cnos.cdd/skills/cdd/epsilon/SKILL.md §2). The cross-cycle aggregator is persistent: .cdd/iterations/INDEX.md accumulates one row per cycle that produced an iteration artifact. The two together form the receipt-stream surface that ε observes — the per-cycle structured findings plus the cross-cycle row-by-row index that lets ε scan many cycles at once.

The failure mode is receipt-stream drift — γ writes the iteration artifact but forgets the aggregator row (ε cannot find the cycle when scanning); γ writes the aggregator row but the per-finding shape diverges from the canonical fields (ε's pattern-detection across cycles breaks); a cross-repo patch-landed finding lands without the bundle trace (the patch's lineage is lost when the source-side bundle archives); the courtesy stub gets misused as "iteration complete" when no actual triage happened (survivorship bias re-emerges). The protocol closes each failure mode by pinning the per-finding shape, the row format, the cadence rule, and the cross-repo bundle invariant explicitly.

This skill is the canonical surface for the receipt-stream wire-format invariants: the per-finding shape inside cdd-iteration.md; the INDEX.md row format; the cadence rule (required when protocol_gap_count > 0; courtesy stub when 0); the finding-disposition vocabulary (patch-landed / next-MCA / no-patch); the cross-repo trace bundle composition. Consumer protocols cite this skill for the wire-format; the role-local authoring procedure (which session writes, what review pass applies) lives in the consumer (cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b is now a pointer; cnos.cdd/skills/cdd/epsilon/SKILL.md §1 is ε's role-local authority on classification).

Authority

• This skill (in cnos.handoff) is the only canonical home for the receipt-stream wire-format invariants: the per-finding shape (Source / Class / Trigger / Description / Root cause / Disposition + per-disposition sub-fields); the INDEX.md row format (eight pipe-separated columns: Cycle | Issue | Date | Findings | Patches | MCAs | No-patch | Path); the cadence rule; the finding-disposition vocabulary; the cross-repo trace bundle invariant. If another skill or doctrine surface carries the same wire-format claims, this skill governs.
• The consumer protocol packages carry the role-local authoring procedure. cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b is a pointer to this skill for the wire-format and applies the rule at cycle close-out (γ writes the artifact per this skill's shape; γ appends the INDEX row). cnos.cdd/skills/cdd/epsilon/SKILL.md §1 is the role-local authority on the CDD-specific gap-class instantiation (cdd-skill-gap / cdd-protocol-gap / cdd-tooling-gap / cdd-metric-gap); the gap-class names are CDD-side. CDR has its own gap-class set per cnos.cdr/skills/cdr/epsilon/SKILL.md §1 (cdr-data-gate-gap / cdr-overclaim-gap / etc.); the per-finding shape and aggregator row format are protocol-agnostic.
• The cross-repo trace bundle composes with cnos.handoff/skills/handoff/cross-repo/SKILL.md. When a finding's disposition is patch-landed with a Cross-repo target, γ creates a bundle at .cdd/iterations/cross-repo/{counterpart-repo}/{slug}/ per the cross-repo skill's case-(b) / case-(c) bundle file set. This skill names the trigger; cross-repo names the bundle composition.

Scope

In scope:

• The per-finding shape inside cdd-iteration.md (F1, F2, … with Source / Class / Trigger / Description / Root cause / Disposition and per-disposition sub-fields: Patch + Affects + Cross-repo for patch-landed; Issue filed + First AC for next-MCA; Reason for no-patch).
• The cross-cycle aggregator (.cdd/iterations/INDEX.md) row format: eight pipe-separated columns naming the cycle / issue / date / findings count / patches count / MCAs count / no-patch count / iteration-file path.
• The cadence rule: per-cycle artifact required when protocol_gap_count > 0; courtesy empty-findings stub permitted (not required) when protocol_gap_count == 0 (cycle/401 convention).
• The finding-disposition vocabulary: patch-landed | next-MCA | no-patch (no other dispositions are valid).
• The aggregator location: .cdd/iterations/INDEX.md at the repo root (the live, persistent aggregator; data, not doctrine — content stays put; doctrine about it lives here).
• The cross-repo trace bundle invariant: when a finding's disposition is patch-landed with Cross-repo target, γ creates a bundle at the canonical path per cnos.handoff/skills/handoff/cross-repo/SKILL.md.
• Empirical anchors: cycle/335 (aggregator initialized); cycle/401 (courtesy stub convention landed); every cycle since cnos#364 carries an INDEX row + a cdd-iteration.md.

Out of scope:

• The role-local authoring procedure (which session writes; the close-out gate ordering; the ε-vs-γ collapse rule). Lives in cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b (cdd-side runbook, now pointer to this skill) and cnos.cdd/skills/cdd/epsilon/SKILL.md §1 (ε's CDD-side role-local authority).
• The CDD-specific gap-class names (cdd-skill-gap / cdd-protocol-gap / cdd-tooling-gap / cdd-metric-gap). Lives in cnos.cdd/skills/cdd/epsilon/SKILL.md §1. CDR has different gap classes; future c-d-X may have other gap classes. The per-finding Class field is consumer-specific in vocabulary; the wire-format Class: slot is protocol-agnostic.
• The receipt schema itself (the typed #Receipt that carries protocol_gap_count and protocol_gap_refs). Lives in schemas/cdd/receipt.cue. This skill consumes the receipt's protocol_gap_count as the cadence-rule input; it does not define the receipt's shape.
• The per-cycle artifact-channel (where the cdd-iteration.md file sits during the cycle; the release-time directory move). Lives in cnos.handoff/skills/handoff/artifact-channel/SKILL.md. The per-cycle iteration artifact is an artifact-channel-owned file; the receipt-stream owns its content shape and the cross-cycle aggregation; the channel owns the file path / write ownership / freeze-on-merge rules.
• The cross-repo bundle composition (LINEAGE.md schemas per case; STATUS state machine; feedback-patch format; archival rule). Lives in cnos.handoff/skills/handoff/cross-repo/SKILL.md. This skill names the trigger for a bundle (a patch-landed finding with Cross-repo target); cross-repo names the bundle.
• The activation-side cadence declaration (per-repo cdd-iteration.md cadence; D/C/B/A severity scale; auto-spawn MCA trigger). Lives in cnos.cdd/skills/cdd/activation/SKILL.md §22. The activation skill declares per-repo cadence overlays; this skill carries the generic wire-format invariants.
• Tooling that auto-verifies receipt-stream discipline (cn cdd verify cycle-iteration checks). Deferred; doctrine first, tooling later.

---

1. Per-finding shape (cdd-iteration.md)

Each finding inside cdd-iteration.md carries the following fields, in this order. The shape is preserved verbatim from the pre-migration canonical home (cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b):

### F1: <title>

- **Source:** α-closeout / β-review / γ-triage / PRA §3 / PRA §4b — name the artifact and section
- **Class:** `cdd-skill-gap` | `cdd-protocol-gap` | `cdd-tooling-gap` | `cdd-metric-gap`
- **Trigger:** §9.1 trigger N | "γ process-gap check" | "review pattern across cycles"
- **Description:** one paragraph
- **Root cause:** one paragraph
- **Disposition:** `patch-landed` | `next-MCA` | `no-patch`

If `patch-landed`:
- **Patch:** `<commit-sha>` or `<file-path>`
- **Affects:** `<cdd skill file or section>` (e.g. `cdd/review/SKILL.md` §3.13)
- **Cross-repo:** `<target-repo>` PR `#NN` if patch landed in a different repo than this cycle (otherwise omit)

If `next-MCA`:
- **Issue filed:** `<repo>` `#NN`
- **First AC:** ...

If `no-patch`:
- **Reason:** required justification per `cnos.cds/skills/cds/CDS.md` §"Closure" → §"Closure rule"

Field meaning:

• Source. Which close-out artifact (α-closeout / β-review / γ-triage / PRA §3 / PRA §4b) surfaced this finding. Names the artifact + section; lets ε trace the finding back to its origin.
• Class. The CDD-side gap class per cnos.cdd/skills/cdd/epsilon/SKILL.md §1 — cdd-skill-gap (procedural skill underspecified / wrong); cdd-protocol-gap (CDD doctrine drifted from invariant); cdd-tooling-gap (tooling absent / wrong / unavailable); cdd-metric-gap (measurement missing / wrong). CDR uses different classes per cnos.cdr/skills/cdr/epsilon/SKILL.md §1; future c-d-X may use other classes.
• Trigger. Which trigger fired — a cnos.cds/skills/cds/CDS.md §"Assessment" → §"Cycle iteration triggers" row (e.g. review rounds > 2; mechanical ratio > 20%; avoidable tooling/environmental failure; loaded skill failed to prevent a finding), or the γ process-gap check, or a multi-cycle pattern surfaced by ε's review-pattern-across-cycles read.
• Description. One paragraph naming what happened, in concrete cycle-local terms.
• Root cause. One paragraph naming why it happened — what protocol invariant was unmet, what skill was underspecified, what tool was missing.
• Disposition. One of three values: patch-landed (the patch shipped in this cycle), next-MCA (issue filed for a follow-on cycle), no-patch (explicit no-patch with required reason). No other dispositions are valid.

Per-disposition sub-fields:

• patch-landed — Patch: names the commit SHA or file path; Affects: names the cdd skill file or section that received the patch; Cross-repo: names the target-repo PR if the patch landed in a different repo (otherwise omit).
• next-MCA — Issue filed: names the repo + issue number; First AC: names the first acceptance criterion the follow-on cycle ships against.
• no-patch — Reason: carries the required justification per cnos.cds/skills/cds/CDS.md §"Closure" → §"Closure rule" (one of: all findings were application gaps with adequate existing spec; zero review findings this cycle; environmental failure with no spec-level fix available).

The shape is not redesigned by this skill; the migration is wholesale. Future tightening (adding fields, changing field semantics) is a follow-on cycle; this skill carries the v0.1 shape.

---

2. Aggregator (.cdd/iterations/INDEX.md)

The cross-cycle aggregator at .cdd/iterations/INDEX.md carries one row per cycle that produced an iteration artifact. The row format is eight pipe-separated columns, preserved verbatim from the pre-migration canonical home:

| Cycle | Issue | Date | Findings | Patches | MCAs | No-patch | Path |
|-------|-------|------|----------|---------|------|----------|------|
| {N}   | #N    | YYYY-MM-DD | M | P | A | Z | .cdd/releases/{X.Y.Z}/{N}/cdd-iteration.md |

Column meanings:

Column	Meaning
Cycle	The cycle identifier (typically the issue number; may include sub-cycle suffix like 344-b for multi-phase docs cycles).
Issue	The GitHub issue number (#NNN). Usually equals the cycle but may differ for follow-on cycles re-opening an earlier issue.
Date	ISO date (YYYY-MM-DD) of cycle close.
Findings	Count of findings inside the cycle's cdd-iteration.md (M). Zero for courtesy stubs (cycle/401 convention).
Patches	Count of findings with disposition patch-landed (P). Includes both same-cycle patches and cross-repo patch-landed shipments.
MCAs	Count of findings with disposition next-MCA (A). Each represents one follow-on cycle scheduled.
No-patch	Count of findings with disposition no-patch (Z). Each carries a recorded Reason.
Path	Relative path to the cycle's cdd-iteration.md artifact. Uses .cdd/releases/{X.Y.Z}/{N}/cdd-iteration.md for version-tagged releases; uses .cdd/releases/docs/{ISO-date}/{N}/cdd-iteration.md for docs-only releases; uses .cdd/unreleased/{N}/cdd-iteration.md for in-flight or unreleased cycles.

Row update procedure:

After writing cdd-iteration.md (or the courtesy empty-findings stub), γ updates .cdd/iterations/INDEX.md with one row appended at the bottom of the data table:

1. Open .cdd/iterations/INDEX.md on the cycle branch.
2. Append one row matching the format above, with the cycle's actual values ({N} / #N / ISO date / M / P / A / Z / path).
3. The four count columns sum: Findings = Patches + MCAs + No-patch (M = P + A + Z). This is a derived consistency check; ε's pattern-detection relies on it.
4. Commit on the cycle branch (same commit as cdd-iteration.md if both are written together).

The aggregator file is data, not doctrine. Its content rows are not moved by this migration; this skill names the format and the update procedure. The live file at .cdd/iterations/INDEX.md is the authoritative aggregator for the cnos repo at the current scale.

---

3. Cadence rules

The per-cycle iteration artifact's cadence is conditional on the cycle's receipt:

3.1. Required: protocol_gap_count > 0

When the cycle's receipt has protocol_gap_count > 0 — i.e. the close-out triage in gamma-closeout.md produced ≥ 1 finding tagged cdd-skill-gap, cdd-protocol-gap, cdd-tooling-gap, or cdd-metric-gap — the per-cycle iteration artifact is required. γ MUST author .cdd/unreleased/{N}/cdd-iteration.md with each finding structured per the per-finding shape in §1; γ MUST append one INDEX.md row per §2; γ closure gate blocks closure until both exist.

The receipt's protocol_gap_count field is consistency-pinned to len(protocol_gap_refs) by schemas/cdd/receipt.cue. The receipt is the always-present record of whether an iteration file is required; the file is the conditional record of what was found. This is the inherited rule from ROLES.md §4b.4.

3.2. Permitted: courtesy empty-findings stub when protocol_gap_count == 0

When the cycle's receipt has protocol_gap_count == 0 — i.e. the cycle ran cleanly with no cdd-*-gap findings — the per-cycle iteration artifact is optional. The cycle's gamma-closeout names the no-gap result; no iteration file is required.

However, γ MAY write a courtesy empty-findings stub at .cdd/unreleased/{N}/cdd-iteration.md for traceability. The stub records protocol_gap_count: 0 explicitly, names the no-findings result, and (optionally) explains why no protocol gap surfaced. The stub variant is the convention landed in cycle/401 (the empirical anchor for the courtesy-stub rule). Backward compatibility: existing empty-findings cdd-iteration.md files written under the prior every-cycle rule remain valid artifacts — the rule is "required when > 0", not "forbidden when == 0".

When a courtesy stub is written, γ MAY also append the INDEX.md row (with Findings = 0; Patches = 0; MCAs = 0; No-patch = 0). This is the dominant cnos cdd convention at current scale: nearly every recent cycle has an INDEX row, even for protocol_gap_count == 0 cycles, because the row supports ε's per-cycle scan-without-omission discipline.

3.3. Activation-side overlay

The per-repo cadence declaration (whether courtesy stubs are mandatory in a given project; the D/C/B/A + info severity scale for findings; the sliding-window auto-spawn MCA trigger) is set by cnos.cdd/skills/cdd/activation/SKILL.md §22. The activation skill overlays per-repo policy on top of the generic wire-format invariants this skill carries; the two are layered.

---

4. Aggregator location

The aggregator file lives at:

.cdd/iterations/INDEX.md

at the repo root. The file is data, not doctrine: its content rows accumulate cycle by cycle and are not moved by this migration. The doctrine about the file (the row format; the per-row update procedure; the eight column meanings) is what this skill carries.

The filename will eventually become consumer-specific. CDS's analogue is <project>/.cds/iterations/INDEX.md per cnos.cds/skills/cds/CDS.md §"Closure". CDR's analogue is the per-project coherence-log per cnos.cdr/skills/cdr/epsilon/SKILL.md §1 and cnos.cdr/docs/empirical-anchor-cph.md row #14. The wire-format pattern (one persistent aggregator at a known per-protocol-package path; one row per cycle that produced an iteration artifact; eight columns naming cycle / issue / date / findings / patches / MCAs / no-patch / path) is what this skill records; the per-protocol filename is consumer-side.

For the cnos repo at the current operating point, the file is at .cdd/iterations/INDEX.md. The bootstrap row was added by cycle/335 (per cnos.cds/docs/empirical-anchor-cdd.md #335 row); every cycle since cnos#364 carries one row.

---

5. Finding-disposition discipline

Each finding carries one of three dispositions; the vocabulary is closed (no other dispositions are valid):

• patch-landed — the patch shipped in this cycle. The finding is closed by the same commit set that produced the cycle's release. The patch lands in a CDD skill file, a doc, a tool, or (rarely) a runtime surface. If the patch landed in a different repo, the Cross-repo: sub-field names the target-repo + PR; the cross-repo trace bundle attaches per §6.
• next-MCA — the patch is deferred to a follow-on cycle. The cycle files an issue (the MCA), records the issue number under Issue filed:, and pins the First AC: the follow-on cycle ships against. This is the dominant disposition for findings that are not safe to ship inside the current cycle (cross-cycle redesigns; surfaces with > 1 dependent cycle).
• no-patch — the finding is closed without a patch. This requires explicit justification per cnos.cds/skills/cds/CDS.md §"Closure" → §"Closure rule": one of (a) all findings were application gaps with adequate existing spec (the skill was right; the agent didn't follow it; patch the operator, not the spec); (b) zero review findings this cycle (the finding is a γ-side observation, not a review finding); (c) environmental failure with no spec-level fix available (tooling constraint; cross-cutting infrastructure). The Reason: sub-field is required; silence is not an acceptable disposition.

The vocabulary is classification-grade: ε's pattern-detection across cycles requires that the three dispositions are mutually exclusive and exhaustive. Adding a fourth disposition (e.g., "drop") is a follow-on cycle; this skill carries the v0.1 vocabulary.

The role-local authority for the disposition discipline is cnos.cdd/skills/cdd/epsilon/SKILL.md §1 (CDD-side). The MCA-discipline shape (ship-now / next-MCA / no-patch) is inherited from ROLES.md §4b.5; the CDD-specific gap-class names (cdd-skill-gap / cdd-protocol-gap / etc.) live in cnos.cdd/skills/cdd/epsilon/SKILL.md §1.

---

6. Cross-repo trace

When a finding's Disposition is patch-landed with a Cross-repo: target — i.e. the patch shipped in a different repo than this cycle — γ creates a cross-repo trace bundle at the canonical path:

.cdd/iterations/cross-repo/{counterpart-repo}/{slug}/

per cnos.handoff/skills/handoff/cross-repo/SKILL.md. The bundle composition depends on the case (per cnos.handoff/skills/handoff/cross-repo/SKILL.md §2 directional cases):

• Outbound iteration trace (case b) — the bundle file set is LINEAGE.md with per-patch confirmation; one outbound destination patch.
• Bilateral iteration (case c) — the bundle additionally carries cdd-iteration.md and one feedback patch per outbound destination patch.

Archival. The source-side (cnos-side, for outbound) bundle MAY be archived once the counterpart-side mirror confirms receipt per cnos.handoff/skills/handoff/cross-repo/SKILL.md §"Bundle archival rule"; target-side mirrors are preserved indefinitely. The asymmetric archival rule preserves the bundle's lineage on the target side past the source-side cleanup.

This skill names the trigger (a patch-landed finding with Cross-repo target); cnos.handoff/skills/handoff/cross-repo/SKILL.md names the bundle's composition and archival rule. The two skills compose: the receipt-stream's per-finding shape exposes the trigger; cross-repo carries the bundle.

---

7. Empirical anchors

Every cycle since cnos#364 carries an INDEX.md row + a cdd-iteration.md (file or courtesy stub). Representative anchors:

• cycle/335 — .cdd/iterations/INDEX.md initialized; bootstrap for the receipt-stream artifact. Per cnos.cds/docs/empirical-anchor-cdd.md #335: "Initialized .cdd/iterations/INDEX.md; bootstrap for the receipt-stream artifact". The 8-column row format crystallized here.
• cycle/401 — the courtesy empty-findings stub convention landed. Before cycle/401, the every-cycle cdd-iteration.md rule produced files for protocol_gap_count == 0 cycles too (the survivorship-bias ambiguity: "is the missing file a clean cycle or a skipped close-out?"). cycle/401 resolved the ambiguity structurally — the receipt's protocol_gap_count field is the always-present record of whether an iteration file is required; the file is the conditional record of what was found. Empty-findings files written under the prior rule remain valid artifacts (backward compatibility); future cycles with protocol_gap_count == 0 may write a courtesy stub or omit the file entirely.
• cycle/369 — Phase 2 receipt schemas; first cycle to record cdd-protocol-gap findings under the typed receipt + iteration shape. 3 findings, 0 patches, 2 MCAs, 1 no-patch.
• cycle/388 — Phase 2.5 schemas split; demonstrated the cdd-protocol-gap + cdd-skill-gap mixed-class triage in one cycle. 3 findings, 0 patches, 2 MCAs.
• cycle/393 — δ-as-architect protocol patches; 4 patches landed on the iteration surface in one cycle (the patch-landed disposition exercised at scale). 0 findings new; 4 patches landed against prior MCAs.
• cycle/410 — Sub 5 of cnos#403 (eight-section migration); closure with courtesy cdd-iteration.md stub (zero findings). Demonstrates the closure-stub option as the standard pattern at current operating point.
• The #406–#412 wave — post-codification cycles that exercised the receipt-stream in steady-state. Every cycle in the wave produced a complete cdd-iteration.md (or courtesy stub) + an INDEX.md row; the receipt-stream wire-format is operationally stable.

The full row history is at .cdd/iterations/INDEX.md — 22+ rows at the source pin, accumulating one per cycle. The empirical anchor index for CDS-class cycles is at cnos.cds/docs/empirical-anchor-cdd.md (which cross-references the receipt-stream's per-cycle artifacts as the canonical receipt instances per CDS Field 3).

---

8. Related documents

• cnos.handoff/skills/handoff/HANDOFF.md — package contract; this skill is one of five sub-surfaces.
• cnos.handoff/skills/handoff/artifact-channel/SKILL.md — the per-cycle channel substrate. The cdd-iteration.md file lives on this channel as a γ-owned file (per artifact-channel §2.2); the channel governs the file path, write ownership, and freeze-on-merge rule; this skill governs the file's content shape and the cross-cycle aggregation.
• cnos.handoff/skills/handoff/cross-repo/SKILL.md — the cross-repo bundle composition. When a patch-landed finding has a Cross-repo target, the bundle attaches per the cross-repo skill's case-(b) / case-(c) bundle file set + asymmetric archival rule.
• cnos.handoff/skills/handoff/dispatch/SKILL.md — the dispatch wire format; the dispatch prompt initiates the cycle whose close-out feeds this receipt-stream.
• cnos.handoff/skills/handoff/mid-flight/SKILL.md — the rescue mechanism; orthogonal to receipt-stream (rescue happens during the cycle; receipt-stream consumes the cycle's close-out artifact).
• cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b — the cdd-side runbook (now a pointer to this skill). post-release/SKILL.md applies the rule at cycle close-out: γ writes cdd-iteration.md per the canonical shape in §1; γ appends the INDEX row per §2; γ runs the cross-repo bundle composition per §6 if applicable.
• cnos.cdd/skills/cdd/epsilon/SKILL.md §1 — ε's CDD-side role-local authority on the gap-class instantiation (cdd-skill-gap / cdd-protocol-gap / cdd-tooling-gap / cdd-metric-gap). The Class: field in this skill's per-finding shape takes one of these four values for CDD-side findings; CDR's Class: values are different per cnos.cdr/skills/cdr/epsilon/SKILL.md §1.
• cnos.cdd/skills/cdd/activation/SKILL.md §22 — the per-repo cadence declaration (whether courtesy stubs are mandatory; D/C/B/A + info severity scale; sliding-window auto-spawn MCA trigger). The activation skill overlays per-repo policy on top of the generic wire-format invariants this skill carries.
• cnos.cdd/skills/cdd/COHERENCE-CELL.md §"ε as Protocol Evolution" + §"ε Artifact Rule" — the cell-doctrinal grounding for ε's role; the receipt-stream is the artifact surface ε reads.
• cnos.cds/skills/cds/CDS.md §"Closure" — CDS-class closure gate; F9 ("Missing cdd-iteration.md when triggers fired") names the gate that blocks closure when protocol_gap_count > 0 but no iteration artifact exists. CDS.md cites this skill (post-#419) for the per-finding shape; the operational realization in cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b is a pointer to this skill.
• cnos.cds/skills/cds/CDS.md §"Artifact contract" → §"Location matrix" — declares the cdd-iteration.md file's canonical path in the CDS-class artifact set; the file lives on the artifact-channel and its content shape is governed by this skill.
• schemas/cdd/receipt.cue (#Receipt) — the typed receipt that carries protocol_gap_count and protocol_gap_refs. This skill consumes the receipt's protocol_gap_count as the cadence-rule input (§3.1).
• ROLES.md §4b — the generic ε doctrine; the authoritative source for the watched-fields invariant, the MCA discipline, and the ε=δ collapse rule. The cadence rule (required only when protocol_gap_count > 0) is inherited from §4b.4.
• cnos#335 — INDEX.md aggregator bootstrap empirical anchor.
• cnos#364 — first cycle in the steady-state INDEX.md every-cycle period.
• cnos#401 — courtesy empty-findings stub convention empirical anchor.
• cnos#404 — parent tracker (handoff/coordination extraction wave).
• cnos#419 — the codification cycle for this surface (Sub 5 of cnos#404).

---

9. Non-goals

• This skill does NOT redesign the per-finding shape. The six top-level fields (Source / Class / Trigger / Description / Root cause / Disposition) and the per-disposition sub-fields are preserved verbatim from the pre-migration canonical home (cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b). Future tightening is a follow-on cycle.
• This skill does NOT redesign the INDEX.md row format. The eight pipe-separated columns (Cycle | Issue | Date | Findings | Patches | MCAs | No-patch | Path) are preserved verbatim. Future tightening (e.g., adding a sub-cycle suffix column, separating same-cycle from cross-repo patch counts) is a follow-on cycle.
• This skill does NOT modify .cdd/iterations/INDEX.md content. The file's existing rows stay where they are; only the standard close-out row for this cycle is added (per the artifact-channel cycle close-out rule and the post-release §5.6b pointer this skill now governs).
• This skill does NOT own the role-local authoring procedure. Which session writes the artifact, the close-out gate ordering, the ε=δ collapse rule for the current operating point — those live in cnos.cdd/skills/cdd/post-release/SKILL.md §5.6b (cdd-side runbook, now pointer) and cnos.cdd/skills/cdd/epsilon/SKILL.md (ε role-local).
• This skill does NOT own the CDD-specific gap-class names (cdd-skill-gap / cdd-protocol-gap / cdd-tooling-gap / cdd-metric-gap). Those live in cnos.cdd/skills/cdd/epsilon/SKILL.md §1. The Class: slot in this skill's per-finding shape is protocol-agnostic; the values are consumer-specific.
• This skill does NOT own the cross-repo bundle composition (LINEAGE.md schemas; STATUS state machine; feedback-patch format; archival rule). Those live in cnos.handoff/skills/handoff/cross-repo/SKILL.md. This skill names the trigger (a patch-landed finding with Cross-repo target); cross-repo names the bundle.
• This skill does NOT own the polling primitives or harness substrate. Those live in cnos.cdd/skills/cdd/harness/SKILL.md §5.4 and cnos.cds/skills/cds/CDS.md §"Coordination surfaces" → §"Polling primitives". This skill is an artifact-surface for a cross-cycle stream; the read mechanism (how ε polls the aggregator across cycles) is harness-side.
• This skill does NOT lift the per-finding shape or the INDEX row format into a typed schema (e.g. schemas/handoff/iteration-finding.cue / schemas/handoff/iteration-index.cue). The Markdown invariants are the v0.1 canonical rendering; a typed schema is deferred per the extraction-map.md §6 note (Sub 3's Q2 ruling precedent — schema stays Markdown for v0.1). Re-evaluate if a downstream consumer (CCNF-O Track A; cn cdd verify; cross-repo bundle validators) needs a typed shape.

---

10. Kata

Scenario

γ has just closed a cycle. β has merged the cycle branch to main; α and β close-outs are on main. γ has filed gamma-closeout.md. The cycle's receipt at .cdd/unreleased/{N}/receipt.yaml carries protocol_gap_count: 2 with two refs naming cdd-skill-gap findings surfaced during review.

Expected reasoning

1. Cadence check (§3.1). The receipt has protocol_gap_count > 0 (specifically, 2). The per-cycle iteration artifact is required, not optional. γ MUST author .cdd/unreleased/{N}/cdd-iteration.md and MUST append one INDEX.md row.

2. Per-finding authoring (§1). γ opens .cdd/unreleased/{N}/cdd-iteration.md and writes one F1 / F2 section per finding. Each section carries the six top-level fields verbatim per the per-finding shape:

   • Source: names the artifact + section (e.g. β-review §3 R1).
   • Class: is cdd-skill-gap (per the receipt's refs).
   • Trigger: names the CDS §"Assessment" trigger that fired (e.g. "mechanical ratio > 20% with ≥ 10 findings" if applicable).
   • Description: one paragraph of cycle-local detail.
   • Root cause: one paragraph naming the unmet invariant.
   • Disposition: one of patch-landed / next-MCA / no-patch. γ picks based on whether the patch shipped this cycle, is deferred to a follow-on, or is closed with explicit no-patch reason.

3. Per-disposition sub-fields (§1). For each finding:

   • If patch-landed: γ records the commit SHA / file path under Patch:, names the affected skill file/section under Affects:, and (if cross-repo) names the target-repo + PR under Cross-repo:.
   • If next-MCA: γ files the issue (e.g. gh issue create), records the issue number under Issue filed:, and pins the First AC:.
   • If no-patch: γ records the Reason: per cnos.cds/skills/cds/CDS.md §"Closure" → §"Closure rule".

4. Aggregator row (§2). γ opens .cdd/iterations/INDEX.md and appends one row:

   | N | #N | YYYY-MM-DD | 2 | P | A | Z | .cdd/unreleased/N/cdd-iteration.md |
   

   where P + A + Z = 2 (the consistency check per §2). γ commits both files in the same commit (typical pattern: the γ-closeout commit also carries the iteration artifact + INDEX row).

5. Cross-repo bundle (§6). If any finding had Disposition: patch-landed with a Cross-repo: target, γ creates a bundle at .cdd/iterations/cross-repo/{counterpart-repo}/{slug}/ per cnos.handoff/skills/handoff/cross-repo/SKILL.md. The bundle's LINEAGE.md names the source-side patch + the destination-side PR.

6. Release-time move (artifact-channel §2.4.2). Before γ requests the tag from δ, γ moves .cdd/unreleased/{N}/ → .cdd/releases/{X.Y.Z}/{N}/ (or .cdd/releases/docs/{ISO-date}/{N}/ for docs-only releases). The cdd-iteration.md file rides the move; the INDEX.md row's Path column is updated to reflect the new path in a subsequent close-out commit if not already pinned at write time.

Common failures

• γ writes cdd-iteration.md but forgets the INDEX row — ε's pattern-detection across cycles skips this cycle when scanning the aggregator. CDS gate F9 catches this.
• γ writes the INDEX row but the per-finding shape diverges (e.g. uses Status: instead of Disposition:; uses a new class name not in the four-value vocabulary). ε's class-by-class scan across cycles breaks.
• γ writes a no-patch disposition without a Reason: — closure rule is violated; the silence is not an acceptable disposition.
• γ writes a courtesy stub (protocol_gap_count == 0) and labels it "iteration complete" with the no-findings result despite there being findings. The survivorship-bias ambiguity re-emerges; ε cannot distinguish clean cycles from skipped triage.
• A patch-landed finding with a Cross-repo: target lands without the bundle at .cdd/iterations/cross-repo/{counterpart-repo}/{slug}/. The patch's lineage is lost when the source-side bundle archives.

Verification

• .cdd/unreleased/{N}/cdd-iteration.md exists if protocol_gap_count > 0.
• .cdd/iterations/INDEX.md has one new row for cycle N with the eight columns populated.
• For each finding, the six top-level fields are present + the per-disposition sub-fields are present.
• Findings = Patches + MCAs + No-patch in the INDEX row (M = P + A + Z).
• For each patch-landed + Cross-repo finding, .cdd/iterations/cross-repo/{counterpart-repo}/{slug}/ exists with the bundle.

Reflection

A second γ reading the cycle's record across cycles (via .cdd/iterations/INDEX.md row-by-row) should be able to surface protocol-gap patterns no single cycle could surface — e.g., "five cycles in a row had cdd-skill-gap findings against review/SKILL.md §3.13; the skill is underspecified and needs a patch". The receipt-stream is the cross-cycle learning signal; without it, ε would have to re-derive cycle context from gh logs and per-cycle assessment files for every triage. The artifact-channel carries the per-cycle record; the receipt-stream carries the cross-cycle pattern surface.

---

End of skill.