[Look-Ahead §5] scenario-forecast long-horizon-mode + forward-indicators multi-horizon decay

[Copilot]

scenario-forecast.md lacked any mechanism to enforce the deeper scenario set required for term/election-cycle runs; forward-indicators.md had no horizon-tagging contract, leaving indicators active beyond their predictive shelf-life.

Template changes

• scenario-forecast.md — YAML frontmatter long-horizon-mode: false flag (moved to byte 0 so the aggregator strips it correctly) + new §0 section specifying the override contract for term-outlook and election-cycle article types:

  • ≥ 6 scenarios (replaces the 3–5 standard cap)
  • Mandatory EP-election outcome branches (centre-right / centre-left / fragmented)
  • Mandatory regime-change branch (unconditional in long-horizon mode)
  • ≥ 2 wildcard/black-swan branches
  • Per-scenario WEP bands drawn from forward-projection-methodology.md §3 (single source of truth — not duplicated here)
  • Anti-patterns table updated: "5+ scenarios" rule explicitly exempted in long-horizon mode
  • Activation signals section accurately describes enforcement: Stage-C gates on manifest.articleType, not registry fields
  • §0 wording clarifies that Stage C enforces the gate regardless of the frontmatter flag for enforced article types

• forward-indicators.md — New §Multi-Horizon Decay Table mapping each indicator class to the horizons where it retains signal (✅) vs. has decayed (⬜), across T+7d / 30d / 90d / 12m / term-end / EP-election. Indicator card template gains a **Horizons:** field (8 structure fields total) so monitoring plans can retire indicators at their decay point. Canonical WEP numbers referenced to forward-projection-methodology.md §3, not duplicated.

Enforcement

reference-quality-thresholds.json — new structuralRequirements.longHorizonScenarioGate:

"longHorizonScenarioGate": {
  "articleTypes": ["term-outlook", "election-cycle"],
  "minScenarios": 6,
  "artifact": "intelligence/scenario-forecast.md"
}

validate-analysis-completeness.js — new functions wired into main():

• getScenarioCountableContent(content) — strips everything after the ## … Worked example H2 so template examples don't satisfy the gate
• countScenarios(content) — counts ### Scenario [A-Za-z0-9]+(?:-[A-Za-z0-9]+)*[:—] headings in the countable region; supports hyphenated IDs (e.g. A-24) while still excluding underscore-containing identifiers to prevent false positives
• validateLongHorizonScenarioGate(runDir, rules) — emits hard-RED long-horizon-scenario-count:N<6 when the gate fires; throws long-horizon-scenario-gate:invalid-config when the config targets the current articleType but artifact/minScenarios are missing or invalid (no silent fallback defaults); includes path-traversal guard rejecting absolute paths and .. segments to prevent reading outside runDir
• WEP_BAND_RE updated to include "About even" so artifacts using the canonical forward-projection-methodology.md §3 label are not incorrectly flagged with wep:missing

Prompt update

11-forward-projection.md §5 expanded with the full long-horizon requirements list, mandatory (unconditional) regime-change branch, and explicit cross-links to scenario-forecast.md §0 and forward-indicators.md §decay-table. No double source of truth: both templates point at forward-projection-methodology.md §3 for WEP bands.

[Copilot]

Pull request overview

Adds long-horizon enforcement and horizon-aware indicator retirement rules to the forward-projection analysis artifacts, tightening Stage-C completeness for term/election-cycle runs.

Changes:

• Introduces a Stage-C “long-horizon scenario count” structural gate (≥6 scenarios) driven by reference-quality-thresholds.json.
• Expands scenario-forecast.md with a long-horizon-mode contract and updates forward-indicators.md with a multi-horizon decay table + per-indicator horizon tagging.
• Updates the forward-projection prompt and adds unit tests covering the new validator gate behavior.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
scripts/validate-analysis-completeness.js	Adds scenario heading counter + long-horizon scenario-count gate wired into Stage C.
analysis/methodologies/reference-quality-thresholds.json	Adds structuralRequirements.longHorizonScenarioGate configuration.
test/unit/validate-analysis-completeness.test.js	Adds unit tests for the new long-horizon scenario-count gate.
analysis/templates/scenario-forecast.md	Adds long-horizon-mode specification and updated scenario requirements.
analysis/templates/forward-indicators.md	Adds multi-horizon decay table and Horizons tagging contract for indicators.
.github/prompts/11-forward-projection.md	Updates Stage-B prompt with long-horizon requirements and cross-links.

[Copilot]

buildRules() claims a malformed structuralRequirements.longHorizonScenarioGate “should not bypass the gate”, but if articleType is targeted and artifact/minScenarios are missing/invalid you currently just leave longHorizonScenarioGate = null, which silently disables enforcement. Consider turning this into a hard Stage-C failure (e.g., emit a RED issue like long-horizon-scenario-gate:invalid-config) when the config is present for the current articleType but invalid, and add a unit test for that case.

Suggested change

	// default them; a malformed config should not bypass the gate.
	if (
	typeof lhGateCfg.artifact === 'string' &&
	lhGateCfg.artifact.length > 0 &&
	typeof lhGateCfg.minScenarios === 'number' &&
	lhGateCfg.minScenarios > 0
	) {
	longHorizonScenarioGate = {
	artifact: lhGateCfg.artifact,
	minScenarios: lhGateCfg.minScenarios,
	};
	}
	// default them; a malformed config must fail Stage C instead of bypassing
	// the gate for targeted article types.
	const hasValidArtifact =
	typeof lhGateCfg.artifact === 'string' && lhGateCfg.artifact.trim().length > 0;
	const hasValidMinScenarios =
	typeof lhGateCfg.minScenarios === 'number' && lhGateCfg.minScenarios > 0;
	if (!hasValidArtifact || !hasValidMinScenarios) {
	throw new Error(
	`long-horizon-scenario-gate:invalid-config articleType=${articleType} ` +
	'structuralRequirements.longHorizonScenarioGate must define a non-empty ' +
	'`artifact` and a positive numeric `minScenarios` for targeted article types.'
	);
	}
	longHorizonScenarioGate = {
	artifact: lhGateCfg.artifact.trim(),
	minScenarios: lhGateCfg.minScenarios,
	};

[Copilot]

The YAML frontmatter is not at byte/line 0 because it’s preceded by the SPDX HTML comments. The aggregator’s stripFrontMatter() only removes frontmatter when the document starts with ---, so this block will leak into aggregated/published HTML. Move the frontmatter to the very top (before the SPDX comments) or remove it and express the flag in-body instead.

Suggested change

	<!-- SPDX-FileCopyrightText: 2024-2026 Hack23 AB -->
	<!-- SPDX-License-Identifier: Apache-2.0 -->
	---
	long-horizon-mode: false
	---
	---
	long-horizon-mode: false
	---
	<!-- SPDX-FileCopyrightText: 2024-2026 Hack23 AB -->
	<!-- SPDX-License-Identifier: Apache-2.0 -->

[Copilot]

The longHorizonScenarioGate.note regex description doesn’t match the implementation: countScenarios() accepts alphanumeric IDs (A–Z, 0–9) and uses a global multiline match, but the note says it only counts \d+ and implies a single match. Please update the note so it accurately reflects the validator’s actual heading pattern.

Suggested change

	"note": "Enforced by scripts/validate-analysis-completeness.js countScenarios(). Counts headings matching /^###\\s+Scenario\\s+\\d+[:—]/m in the artifact. Activated for articleTypes listed above (electoralOverlay===true OR scenarioMaxHorizonMonths>=36 in article-horizons registry)."
	"note": "Enforced by scripts/validate-analysis-completeness.js countScenarios(). Counts all headings matching /^###\\s+Scenario\\s+[A-Z0-9]+[:—]/gm in the artifact. Activated for articleTypes listed above (electoralOverlay===true OR scenarioMaxHorizonMonths>=36 in article-horizons registry)."

[Copilot]

The template now adds a required **Horizons:** field, but the checklist still says “all 7 structure fields (including **Horizons:**)”. With Horizons added, the card template appears to have 8 fields (Category, Horizons, Description, Current State, Thresholds, Monitoring Method, Scenario Linkage, Historical Context). Update the checklist (and any earlier “7 fields” wording) so the required field count is consistent with the template.

Suggested change

	- [ ] Each indicator has all 7 structure fields (including `**Horizons:**` tag)
	- [ ] Multi-Horizon Decay Table present in §Multi-Horizon Decay Table
	- [ ] Every indicator tagged with its active horizons (7d / 30d / 90d / 12m / term-end / EP-election)
	- [ ] Each indicator has all 8 structure fields (including required `**Horizons:**`)
	- [ ] Multi-Horizon Decay Table present in §Multi-Horizon Decay Table
	- [ ] Every indicator has the `**Horizons:**` field populated with its active horizons (7d / 30d / 90d / 12m / term-end / EP-election)

[Copilot]

This bullet makes the regime-change branch conditional (“when any tripwire from §2 is armed”), but scenario-forecast.md §0 defines a regime-change branch as mandatory in long-horizon mode. Align the prompt wording with the template/spec so agents don’t follow conflicting requirements.

Suggested change

	- At least **1 regime-change branch** when any tripwire from §2 is armed.
	- Mandatory **regime-change branch**; use tripwires from §2 to define and monitor the branch.

[pethers]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

All five review suggestions applied in commit 6df6f00:

1. Frontmatter position — moved ---\nlong-horizon-mode: false\n--- to byte 0 (before SPDX comments) so cleanArtifact() strips it correctly.
2. Invalid-config error — buildRules() now throws long-horizon-scenario-gate:invalid-config when the gate targets the current articleType but artifact/minScenarios are missing or invalid; main() catches it, emits a RED gate line, and exits 1. Two new unit tests cover the missing-artifact and zero-minScenarios cases.
3. Note regex accuracy — updated reference-quality-thresholds.json note to /^###\s+Scenario\s+[A-Z0-9]+[:—]/gm matching the actual countScenarios() implementation.
4. 8-field checklist — forward-indicators.md checklist updated from "7 structure fields" to "8 structure fields" with the **Horizons:** field explicitly listed.
5. Mandatory regime-change branch — 11-forward-projection.md §5 now reads "Mandatory regime-change branch; use tripwires from §2 to define and monitor the branch." (unconditional, aligned with scenario-forecast.md §0).

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.

[Copilot]

The text implies Stage C activation is based on the article-horizons registry fields (electoralOverlay / scenarioMaxHorizonMonths), but the validator code actually only gates by manifest.articleType (via structuralRequirements.longHorizonScenarioGate.articleTypes). Please reword this section to reflect the actual enforcement mechanism to avoid misleading authors.

[Copilot]

This paragraph states the validator enforces the ≥6 scenario floor for term-outlook/election-cycle “i.e.” any type with electoralOverlay === true OR scenarioMaxHorizonMonths >= 36. The validator doesn’t inspect those registry fields, so the “i.e.” equivalence is inaccurate. Suggest changing to: enforced when manifest.articleType is one of the configured types; the registry fields are the rationale for which types are listed.

Suggested change

	The validator enforces the ≥ 6 scenario floor when the manifest's `articleType` is one of: **`term-outlook`**, **`election-cycle`** (i.e. any type whose registry entry has `electoralOverlay === true` OR `scenarioMaxHorizonMonths >= 36`).
	The validator enforces the ≥ 6 scenario floor when the manifest's `articleType` is one of the configured types: **`term-outlook`**, **`election-cycle`**. Registry fields such as `electoralOverlay === true` and `scenarioMaxHorizonMonths >= 36` explain why those types are included, but are not the fields the validator directly inspects.

[Copilot]

The metadata row marks Long-Horizon Mode as required “when electoralOverlay=true or scenarioMaxHorizonMonths≥36”, but Stage C currently doesn’t validate this flag (it only checks manifest.articleType). To prevent false expectations, either align the wording with the real validator behavior or add a validator check for the frontmatter flag if that’s intended to become enforceable.

[Copilot]

The checklist now asserts “8 structure fields” due to the new **Horizons:** tag, but earlier in the template Section 2 still says “≥10 indicators with all 7 structure fields”. Please update the earlier requirement text and/or the master table header guidance so the template is internally consistent about the expected field count.

[pethers]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

All four review comments addressed in commit 9b63272:

1. §0 flag description (line 18) — removed the registry-field clause; description now simply states the flag applies to term-outlook and election-cycle runs.
2. Activation signals paragraph (line 33) — replaced the inaccurate "i.e." equivalence with the suggested wording: validator gates on manifest.articleType; registry fields (electoralOverlay, scenarioMaxHorizonMonths) explain why those types are listed but are not directly inspected.
3. Metadata row (line 60) — label changed from "when electoralOverlay=true or scenarioMaxHorizonMonths≥36" to "for term-outlook and election-cycle articleTypes" to match the actual gate behavior.
4. Section 2 field count (forward-indicators.md line 88) — updated from "all 7 structure fields" to "all 8 structure fields" for internal consistency with the checklist.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

[Copilot]

countScenarios() only counts headings where the scenario identifier is strictly [A-Za-z0-9]+. In existing artifacts, scenario IDs sometimes contain hyphens (e.g. ### Scenario A-24 — ... in analysis/daily runs), which this regex will not count and could cause false REDs for long-horizon runs if that style is used. Consider allowing hyphens in the identifier (while still excluding underscores) or explicitly documenting/enforcing that hyphens are not permitted in scenario IDs for gated artifacts.

Suggested change

	* and letter variants used in worked examples). Only alphanumeric identifiers
	* are matched to avoid false positives from underscore-containing headings.
	*/
	function countScenarios(content) {
	// Match "### Scenario 1:" or "### Scenario A —" (digit or letter identifier only)
	const re = /^###\s+Scenario\s+[A-Za-z0-9]+\s*[:—]/gm;
	* and letter variants used in worked examples), and also supports
	* hyphen-separated alphanumeric identifiers such as `A-24`. Underscore-
	* containing identifiers are still excluded to avoid false positives.
	*/
	function countScenarios(content) {
	// Match "### Scenario 1:", "### Scenario A —", or "### Scenario A-24 —"
	// while rejecting underscore-containing identifiers.
	const re = /^###\s+Scenario\s+[A-Za-z0-9]+(?:-[A-Za-z0-9]+)*\s*[:—]/gm;

[Copilot]

This template pushes authors to use WEP bands from forward-projection-methodology.md §3, which includes the label “About even”. The Stage-C validator’s WEP_BAND_RE currently matches “Roughly Even”/“Even Chance” but not “About even”, so scenario-forecast artifacts that follow the canonical table can be incorrectly flagged with wep:missing. To keep the “single source of truth” contract, the validator’s WEP matcher should accept “About even” (or the canonical label set should be made consistent).

[pethers]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

Both issues fixed in commit bd6c381:

1. countScenarios() hyphenated IDs — regex updated to /^###\s+Scenario\s+[A-Za-z0-9]+(?:-[A-Za-z0-9]+)*\s*[:—]/gm so IDs like A-24 are counted; underscores still excluded. New unit test 'counts hyphenated scenario IDs (e.g. A-24) correctly' added and passing.

2. WEP_BAND_RE missing "About even" — "About even" added to the matcher so artifacts using the canonical forward-projection-methodology.md §3 label are no longer incorrectly flagged with wep:missing.

reference-quality-thresholds.json note also updated to reflect the hyphen-supporting pattern.

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 5 comments.

[Copilot]

validateLongHorizonScenarioGate() builds absPath via path.join(runDir, artifact). If artifact is absolute (or contains .. segments), path.join can escape runDir and read arbitrary files on the runner. Even though the config is repo-controlled, it’s safer to enforce that artifact is a normalized relative path under runDir (e.g., reject path.isAbsolute(artifact) and verify path.resolve(runDir, artifact) starts with the resolved runDir prefix).

[Copilot]

countScenarios() currently counts every matching ### Scenario ... heading in the entire file. Since analysis/templates/scenario-forecast.md contains a worked example with ### Scenario A/B/C — ... headings, an analysis run can meet the ≥6-scenario gate by keeping the worked example plus only writing 3 real scenarios. To make the gate enforce the intended scenario set, consider excluding the worked-example section (e.g., only count headings before the "## 🛠️ Worked example" H2, or ignore matches that occur after that header).

Suggested change

	* Count the number of scenario headings in a scenario-forecast artifact.
	* Matches `### Scenario N:` and `### Scenario N —` patterns (both numeric
	* and letter variants used in worked examples), and also supports
	* hyphen-separated alphanumeric identifiers such as `A-24`. Underscore-
	* containing identifiers are still excluded to avoid false positives.
	*/
	function countScenarios(content) {
	// Match "### Scenario 1:", "### Scenario A —", or "### Scenario A-24 —"
	// while rejecting underscore-containing identifiers.
	const re = /^###\s+Scenario\s+[A-Za-z0-9]+(?:-[A-Za-z0-9]+)*\s*[:—]/gm;
	const matches = content.match(re);
	* Return the portion of a scenario-forecast artifact that should contribute
	* to the scenario-count gate.
	*
	* Worked examples in the template may contain `### Scenario ...` headings
	* that must not satisfy the minimum authored-scenarios requirement, so
	* everything from the first `## ... Worked example` H2 onwards is excluded.
	*/
	function getScenarioCountableContent(content) {
	const workedExampleHeader = /^##\s+.*Worked example\b.*$/im;
	const match = workedExampleHeader.exec(content);
	if (!match || typeof match.index !== 'number') {
	return content;
	}
	return content.slice(0, match.index);
	}
	/**
	* Count the number of scenario headings in a scenario-forecast artifact.
	* Matches authored `### Scenario N:` and `### Scenario N —` patterns and
	* also supports hyphen-separated alphanumeric identifiers such as `A-24`.
	* Underscore-containing identifiers are still excluded to avoid false
	* positives. Headings in the worked-example section are ignored.
	*/
	function countScenarios(content) {
	const countableContent = getScenarioCountableContent(content);
	// Match "### Scenario 1:", "### Scenario A —", or "### Scenario A-24 —"
	// while rejecting underscore-containing identifiers.
	const re = /^###\s+Scenario\s+[A-Za-z0-9]+(?:-[A-Za-z0-9]+)*\s*[:—]/gm;
	const matches = countableContent.match(re);

[Copilot]

This section says long-horizon-mode is optional and that when it’s unset/false, standard behavior is preserved. However, the new Stage-C gate is enforced based on manifest.articleType (term-outlook/election-cycle) regardless of this frontmatter flag. Consider rewording to avoid implying that a term-outlook/election-cycle run can stay in “standard mode” by leaving the flag false, since Stage C will still require ≥6 scenarios.

Suggested change

	**`long-horizon-mode`** is an optional frontmatter flag. Set it to `true` when writing a `term-outlook` or `election-cycle` analysis run. When unset or `false`, standard behaviour (3–5 scenarios, 7/30/90-day horizon) is preserved.
	### When `long-horizon-mode: true` is set the following requirements override the standard rules:
	**`long-horizon-mode`** is an optional frontmatter flag that marks a run as long-horizon in the template. Set it to `true` when writing a `term-outlook` or `election-cycle` analysis run. However, for validator-enforced long-horizon article types — **`term-outlook`** and **`election-cycle`** — Stage C still applies the long-horizon rules from `manifest.articleType` even if this flag is unset or `false`. Standard behaviour (3–5 scenarios, 7/30/90-day horizon) is preserved only for article types outside that enforced set.
	### For long-horizon runs — whether triggered by `long-horizon-mode: true` or by validator-enforced `manifest.articleType` — the following requirements override the standard rules:

[Copilot]

The parenthetical here claims long-horizon mode is “activated when electoralOverlay === true OR scenarioMaxHorizonMonths >= 36”, but the validator you added enforces the scenario-count gate based on manifest.articleType only (and does not inspect registry fields). To keep the template accurate and avoid confusing editors, update this activation description to match the actual enforcement trigger.

Suggested change

	**Long-horizon mode (`long-horizon-mode: true`, activated when `electoralOverlay === true` OR `scenarioMaxHorizonMonths >= 36`):**
	**Long-horizon mode (`long-horizon-mode: true`, with Stage-C scenario-count enforcement for `term-outlook` or `election-cycle` article types):**

[Copilot]

The structuralRequirements note says the long-horizon scenario gate is “activated when electoralOverlay===true or scenarioMaxHorizonMonths>=36”, but the implementation in scripts/validate-analysis-completeness.js activates it when manifest.articleType is in longHorizonScenarioGate.articleTypes. Suggest updating this note to reflect the actual activation criteria (manifest articleType), or readers may assume registry fields affect enforcement.

Suggested change

"note": "Structural contracts enforced by scripts/validate-analysis-completeness.js. mermaidRequired = artifacts that MUST contain at least one Mermaid block (in addition to the directory-based default for intelligence/, classification/, risk-scoring/, threat-assessment/). readerBlockRequired = artifacts that MUST contain a Reader Briefing / 'For Citizens' / 'Plain Language' section. sourceDiversityRequired = artifacts that MUST cite at least one EP MCP tool reference OR carry an evidence table (markdown table with a Source / Evidence / Reference header column, e.g. a 'Data Sources & Provenance' block); the validator accepts either signal. requiredSections = artifact -> array of H2 substrings that MUST appear (case-insensitive substring match). Added 2026-04 alongside template v2.0 enrichment of impact-matrix, forces-analysis, actor-mapping, political-capital-risk, actor-threat-profiles, legislative-disruption, legislative-velocity-risk, consequence-trees. longHorizonScenarioGate = article types whose intelligence/scenario-forecast.md MUST contain >= 6 scenario headings (### Scenario N pattern); activated when electoralOverlay===true or scenarioMaxHorizonMonths>=36. Added 2026-05 alongside scenario-forecast.md template v1.3.",

"note": "Structural contracts enforced by scripts/validate-analysis-completeness.js. mermaidRequired = artifacts that MUST contain at least one Mermaid block (in addition to the directory-based default for intelligence/, classification/, risk-scoring/, threat-assessment/). readerBlockRequired = artifacts that MUST contain a Reader Briefing / 'For Citizens' / 'Plain Language' section. sourceDiversityRequired = artifacts that MUST cite at least one EP MCP tool reference OR carry an evidence table (markdown table with a Source / Evidence / Reference header column, e.g. a 'Data Sources & Provenance' block); the validator accepts either signal. requiredSections = artifact -> array of H2 substrings that MUST appear (case-insensitive substring match). Added 2026-04 alongside template v2.0 enrichment of impact-matrix, forces-analysis, actor-mapping, political-capital-risk, actor-threat-profiles, legislative-disruption, legislative-velocity-risk, consequence-trees. longHorizonScenarioGate = article types whose intelligence/scenario-forecast.md MUST contain >= 6 scenario headings (### Scenario N pattern); activated for manifest.articleType values listed in longHorizonScenarioGate.articleTypes. Added 2026-05 alongside scenario-forecast.md template v1.3.",

[pethers]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

All five review comments addressed in commit f1acdf9:

1. Path-traversal guard — validateLongHorizonScenarioGate() now rejects absolute paths and .. segments, verifying the resolved path stays under runDir.
2. Worked-example exclusion — new getScenarioCountableContent() strips everything after ## … Worked example H2 before counting scenarios, preventing template examples from satisfying the gate.
3. §0 frontmatter flag wording (line 18-20) — reworded to clarify Stage C enforces via manifest.articleType regardless of the long-horizon-mode flag.
4. Line 323 activation description — parenthetical updated to "with Stage-C scenario-count enforcement for term-outlook or election-cycle article types".
5. reference-quality-thresholds.json note — updated to "activated for manifest.articleType values listed in longHorizonScenarioGate.articleTypes".

Two new unit tests cover the worked-example exclusion and path-traversal rejection (59 tests passing).

[Copilot]

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 3 comments.

[Copilot]

The path-traversal test currently accepts an unhandled exception (stack trace) as a valid failure mode. Once the validator is updated to handle long-horizon gate config errors gracefully, this test should also assert that stdout still contains a STAGE_C_GATE: RED line (so downstream tooling keeps working even on invalid config).

Suggested change

	expect(result.code).toBe(1);
	expect(result.code).toBe(1);
	expect(result.stdout).toContain('STAGE_C_GATE: RED');

[Copilot]

mergeForwardRegistryResult is now used to merge non-forward-registry synthetic results (the long-horizon scenario gate). Renaming this helper to something artifact-agnostic (and updating call sites) would reduce confusion for future maintainers.

[Copilot]

validateLongHorizonScenarioGate can throw (e.g. path traversal / resolves outside runDir), but this call site doesn't catch those errors. That will terminate the validator with an unhandled exception and can prevent emitting the expected STAGE_C_GATE line on stdout, breaking CI/parsers. Wrap this gate invocation in a try/catch (or validate/throw earlier inside buildRules so main’s existing catch handles it) and convert failures into a normal RED result + gate line output.

Suggested change

	const longHorizonScenarioResult = validateLongHorizonScenarioGate(runDir, rules);
	if (longHorizonScenarioResult) {
	mergeForwardRegistryResult(results, longHorizonScenarioResult);
	try {
	const longHorizonScenarioResult = validateLongHorizonScenarioGate(runDir, rules);
	if (longHorizonScenarioResult) {
	mergeForwardRegistryResult(results, longHorizonScenarioResult);
	}
	} catch (error) {
	const message =
	error instanceof Error ? error.message : String(error);
	mergeForwardRegistryResult(results, {
	ok: false,
	ruleId: 'long-horizon-scenario-gate',
	relativePath: 'scenario-forecast.md',
	message: `Long-horizon scenario gate validation failed: ${message}`,
	});

[pethers]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

All three review comments addressed in commit 7eb8442:

1. Renamed helper — mergeForwardRegistryResult → mergeSyntheticResult since it now handles non-forward-registry results (long-horizon gate).
2. try/catch wrapper — validateLongHorizonScenarioGate() call site now catches thrown errors (path traversal, resolve outside runDir) and converts them to a normal RED result with STAGE_C_GATE: RED on stdout instead of an unhandled exception.
3. Path-traversal test — assertion updated to expect(result.stdout).toContain('STAGE_C_GATE: RED') so downstream CI parsers are verified.

All 59 unit tests passing.