@inceptionstack/pi-hard-no 1.0.2 → 1.0.3

package/architect.ts

@@ -44,11 +44,16 @@ You have tools available (read, bash, grep, find, ls) to explore the full codeba
 - Do public APIs have adequate comments/types?
 - Are new files/modules properly documented?
 
-## Response format
-If everything looks good at the big-picture level, say "LGTM — architecture looks solid."
-If there are issues, list them as bullet points with severity (high/medium/low).
-Focus on systemic issues that individual mini-reviews would miss.
-Do NOT repeat issues that were already found and fixed in mini-reviews.`;
+## Response format — terse, technical, low-bandwidth
+
+LGTM or bullets. Nothing else. Agent reads this, not a human — minimum words, maximum signal.
+
+- If everything is fine: one line — "LGTM — architecture looks solid."
+- Otherwise: bullet per issue, each: **<severity>:** <area/file> — <problem>; <fix>.
+- Max 3 lines per issue, 1 line preferred.
+- No preamble, no recap of what you checked, no praise.
+- Skip anything already fixed in mini-reviews.
+- Technical shorthand OK (race, coupling, leak, god-object, dead code).`;
 
 /**
  * Load architect review rules from .hardno/architect.md.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@inceptionstack/pi-hard-no",
-  "version": "1.0.2",
+  "version": "1.0.3",
   "type": "module",
   "description": "Pi extension — automatic code review after every agent turn",
   "license": "MIT",

package/prompt.ts

@@ -70,21 +70,33 @@ export const DEFAULT_AUTO_REVIEW_RULES = `## What to review (in priority order)
 
 // ── Part 3: Suffix (always included, not user-editable) ──
 
-export const PROMPT_SUFFIX = `## Response format
+export const PROMPT_SUFFIX = `## Output style — terse, technical, low-bandwidth
+
+Write as if on a slow link: minimum words, maximum signal. The agent reads this,
+not a human skimming prose.
+
+- Max 3 lines per issue. One line if location + problem + fix fits in one.
+- Each issue = location · problem · fix. Nothing else.
+- No preamble, no recap, no "I noticed", no "consider", no hedging.
+- Technical shorthand OK (e.g. "OBO", "race", "null deref", "unbounded loop").
+- Don't restate code — point to it by file:line.
+- Don't explain why the rule exists. Just the hit.
+
+## Response format
 
 Your response MUST follow this exact structure:
 
-1. (If issues found) List of bullet points, each: - **<Severity>:** <file/location> — <one-line explanation>
+1. (If issues found) List of bullet points, each: - **<Severity>:** <file/location> — <problem>; <fix>
    Severity is one of: High, Medium, Low.
 2. (If no issues) Write a single line: No issues found.
 3. On the final line of your response, output exactly ONE of these verdict tags:
    - <verdict>LGTM</verdict>  — if no real bugs were found
    - <verdict>ISSUES_FOUND</verdict>  — if you flagged any issue above
 
-## Example — issues found
+## Example — issues found (terse)
 
-    - **High:** test-bugs.ts:12 — Off-by-one error: i <= items.length should be i < items.length.
-    - **High:** test-bugs.ts:6 — Hardcoded API key sk-prod-... leaks a secret.
+    - **High:** test-bugs.ts:12 — OBO in \`for (i<=len)\`; use \`<\`.
+    - **High:** test-bugs.ts:6 — hardcoded key \`sk-prod-…\`; move to env.
 
     <verdict>ISSUES_FOUND</verdict>