fix: encode honors context-vs-input governance distinction

[klappy]

Closes the gap between governance and code surfaced during PR #96 testing.

The Gap

Governance (klappy://odd/encoding-types/how-to-write-encoding-types#context-vs-input):

input generates artifacts; context only informs quality scoring

Server code (after PR #96):

const fullInput = context ? `${input}\n${context}` : input;
const artifacts = parseStructuredInput(fullInput, types); // wrong

Result: context paragraphs became separate standalone artifacts instead of informing the primary artifact's score.

The Fix

• runEncodeAction: parsers receive input only (not fullInput)
• runEncodeAction: scoring receives input + context per artifact
• scoreArtifactQuality: accepts optional scoringText parameter
• Inline comments cite the governance doc to prevent regression

Verification

TypeScript compiles clean. Manual verification on preview after merge.

---

Note

Medium Risk
Behavior of encode output and quality ratings changes, which could affect downstream expectations/tests, but the change is localized and not security-sensitive.

Overview
Fixes encode to honor the governance rule that context should not generate artifacts.

runEncodeAction now parses artifacts from input only, and passes context only into quality scoring by appending it to each artifact’s body at scoring time. scoreArtifactQuality is updated to accept optional scoringText, so length/rationale/concreteness checks can consider body + context without changing the persisted artifact content.

^{Reviewed by Cursor Bugbot for commit e95af36. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	oddkit	e95af36	Commit Preview URL Branch Preview URL	Apr 16 2026, 03:11 AM

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Non-empty check passes for empty artifacts with context
  • Changed the non-empty criterion check on line 428 from text.length > 0 back to artifact.body.length > 0 so that context cannot substitute for actual artifact content.

Preview (e95af368db)

diff --git a/workers/src/orchestrate.ts b/workers/src/orchestrate.ts
--- a/workers/src/orchestrate.ts
+++ b/workers/src/orchestrate.ts
@@ -402,15 +402,22 @@
 function scoreArtifactQuality(
   artifact: ParsedArtifact,
   criteria: Array<{ criterion: string; check: string; gapMessage: string }>,
+  scoringText?: string,
 ): { score: number; maxScore: number; level: string; gaps: string[]; suggestions: string[] } {
   const gaps: string[] = [];
   const suggestions: string[] = [];
   let score = 0;
+  // Governance: context informs quality scoring. When scoringText is provided
+  // (artifact.body + context), criteria check against that combined text so
+  // background information in context (rationale, alternatives, evidence)
+  // counts toward the artifact's quality without becoming separate artifacts.
+  // See: klappy://odd/encoding-types/how-to-write-encoding-types#context-vs-input
+  const text = scoringText ?? artifact.body;
 
   if (criteria.length === 0) {
-    if (artifact.body.split(/\s+/).length >= 10) score++;
+    if (text.split(/\s+/).length >= 10) score++;
     else suggestions.push("Expand — more detail improves quality");
-    if (/because|due to|since/i.test(artifact.body)) score++;
+    if (/because|due to|since/i.test(text)) score++;
     else suggestions.push("Add rationale");
     return { score, maxScore: 2, level: score >= 2 ? "adequate" : "weak", gaps, suggestions };
   }
@@ -419,11 +426,11 @@
     const ck = c.check.toLowerCase();
     let passed = false;
     if (ck.includes("non-empty")) passed = artifact.fields.length > 3 || artifact.body.length > 0;
-    else if (ck.includes("10")) passed = artifact.body.split(/\s+/).length >= 10;
-    else if (ck.includes("number") || ck.includes("concrete")) passed = /\d/.test(artifact.body);
+    else if (ck.includes("10")) passed = text.split(/\s+/).length >= 10;
+    else if (ck.includes("number") || ck.includes("concrete")) passed = /\d/.test(text);
     else if (ck.includes("interpretation") || ck.includes("does not contain")) passed = !/should|better|worse|means|implies/i.test(artifact.body);
     else if (ck.includes("prohibition") || ck.includes("requirement")) passed = /must|must not|never|always|shall/i.test(artifact.body);
-    else passed = artifact.body.split(/\s+/).length >= 5;
+    else passed = text.split(/\s+/).length >= 5;
     if (passed) score++;
     else { gaps.push(c.gapMessage); suggestions.push(c.gapMessage); }
   }
@@ -1415,19 +1422,26 @@
   state?: OddkitState,
 ): Promise<ActionResult> {
   const startMs = Date.now();
-  const fullInput = context ? `${input}\n${context}` : input;
+  // Governance: input generates artifacts; context only informs quality scoring.
+  // See: klappy://odd/encoding-types/how-to-write-encoding-types#context-vs-input
+  // Do not pass fullInput to parsers — that would create separate artifacts
+  // for each context paragraph instead of letting context inform scoring.
 
   const types = await discoverEncodingTypes(fetcher, canonUrl);
-  const structured = isStructuredInput(fullInput);
+  const structured = isStructuredInput(input);
   const artifacts = structured
-    ? parseStructuredInput(fullInput, types)
-    : parseUnstructuredInput(fullInput, types);
+    ? parseStructuredInput(input, types)
+    : parseUnstructuredInput(input, types);
 
-  // Score each artifact using its type's quality criteria
+  // Score each artifact using its type's quality criteria.
+  // When context is provided, append it to the artifact's body for scoring
+  // so background information (rationale, alternatives, evidence) counts
+  // toward the artifact's quality without becoming separate artifacts.
   const scoredArtifacts = artifacts.map((a) => {
     const typeDef = types.find((t) => t.letter === a.type);
     const criteria = typeDef ? typeDef.qualityCriteria : [];
-    const quality = scoreArtifactQuality(a, criteria);
+    const scoringText = context ? `${a.body}\n${context}` : undefined;
+    const quality = scoreArtifactQuality(a, criteria, scoringText);
     return { title: a.title, type: a.type, typeName: a.typeName, content: a.body, fields: a.fields, quality };
   });

_{You can send follow-ups to the cloud agent here.}

^{Reviewed by Cursor Bugbot for commit e6e9e6c. Configure here.}