feat(agent): tighten prompt to bound beyond-diff reads + de-generalize cost-model labels

Three changes from operator feedback:

1. Per-comment � attribution restored on inline comments (operator wants
   it back — the PR-level collapsible is collapsed by default, so the
   attribution is the visible signal of per-finding cost share).
   Hidden only when no _tok_attrib was computed (legacy callers / ollama
   path without usage metering).

2. Agent prompt now bounds reads beyond the diff — the single biggest
   driver of input-token bloat on long agent loops:
     * ≤ 5 file reads beyond the diff for the entire review
     * ≤ 80 lines per read (use --offset + --limit)
     * ≤ 3 grep calls beyond the diff (prefer rtk grep)
     * no re-reads of files already seen
     * no directory walks (ls -R, find .)
     * honor .pr-review.json:exclude_paths

3. De-generalize cost_model calibration labels. The OBSERVED_RUNS list
   referred to `gitea_admin/pragent#7` — a real internal repo path that
   blocks commercialization. Replaced with `internal/hardening-PR (16
   files, 1020 insertions / 91 deletions)`. The numbers (input/output
   tokens, steps, duration) are unchanged — only the labels are
   generic.

Tests:
  * test_inline_comment_body_with_attribution_line — asserts 🪙 line
    shows when _tok_attrib is set
  * test_inline_comment_body_no_attribution_no_coin_line — still
    verifies the line is hidden when no attribution data
  * test_observed_report_prices_every_model — asserts no internal
    repo name appears in the rendered report
Co-Authored-By: Claude <noreply@anthropic.com>

.opencode/agents/pragent.md

@@ -85,14 +85,24 @@ read the full file around a flagged line, not just the diff hunk.
    lists the changed files explicitly under "Changed files" — use that as your
    focus list.
 
-3. **Ground findings in context.** For each changed file, before finalizing any
+3. **Ground findings in context — but stay bounded.** For each changed file,
-   finding, `read`/`grep` its **callers, imports, sibling functions, and type
+   before finalizing any finding, `read`/`grep` its **callers, imports, sibling
-   definitions** so your findings reflect how the change is actually used, not
+   functions, and type definitions** so your findings reflect how the change
-   the hunk in isolation. The repo is checked out at the head sha, so the
+   is actually used, not the hunk in isolation. The repo is checked out at the
-   surrounding code is on disk — use it. Keep it bounded: stop exploring a file
+   head sha, so the surrounding code is on disk — use it.
-   once the finding is grounded (1–3 related files per finding); do NOT do
+
-   unbounded whole-repo walks (token cost, and the focus is the diff's
+   HARD budget on reads beyond the diff (this is the single biggest driver of
-   neighbourhood).
+   token cost on long agent loops):
+     * ≤ 5 file reads BEYOND the diff for the entire review. Count them.
+     * ≤ 80 lines per `read` call — use `read --offset N --limit 80` to slice
+       large files; never `cat` a whole 1000-line file.
+     * ≤ 3 grep calls beyond the diff (use `rtk grep` if available; `grep -n`
+       with a precise pattern otherwise).
+     * Do NOT re-read a file you've already seen. The diff is the source of
+       truth — re-reads only confirm what you already know.
+     * Do NOT walk directories (`ls -R`, `find .`) — list explicitly.
+     * Honour `.pr-review.json:exclude_paths` — those files do not exist for
+       you; do not read them even if they appear in the diff.
 
 4. **Run the repo's own checks via bash.** Detect tooling and run it on the
    CHANGED files only (keep it fast, keep tokens low):

pilot/ai_review.py

@@ -1002,8 +1002,10 @@ def inline_comment_body(f: dict) -> str:
         produced replacement code. Language-tagged fences are reserved for
         cross-file patterns the suggestion block can't carry.
       * Reference as a Markdown hyperlink (``[label](url)``) — never a raw URL.
-      * No per-comment token attribution: the PR-level collapsible carries
+      * Per-comment attributed output tokens (`🪙 ~N tok (P% · attributed)`)
-        all telemetry; inline comments stay focused on the code.
+        when the caller passed `compute_attribution` data. Hidden when the
+        finding has no attributed tokens (e.g. legacy callers / ollama path
+        without usage metering).
     """
     badge = _severity_badge(f.get("severity", "medium"))
     body = f"{badge} {f.get('problem', '').strip()}"
@@ -1019,6 +1021,10 @@ def inline_comment_body(f: dict) -> str:
     ref_md = _format_reference(f.get("reference", ""))
     if ref_md:
         body += f"\n\n🔗 **Reference:** {ref_md}"
+    tok = f.get("_tok_attrib")
+    if tok is not None:
+        pct = (f.get("_tok_pct", 0.0) or 0.0) * 100
+        body += f"\n\n🪙 ~{tok} tok ({pct:.0f}% · attributed output)"
     return body
 
 

pilot/cost_model.py

@@ -176,7 +176,7 @@ DEFAULT_TIERS = [
 # from a guess, and the first entry corrected the tier assumptions by ~15x.
 OBSERVED_RUNS: list[dict] = [
     {
-        "label": "gitea_admin/pragent#7 (the hardening PR)",
+        "label": "internal/hardening-PR (16 files, 1020 insertions / 91 deletions)",
         "date": "2026-08-18",
         "tier": "full",
         "diff_tokens": 17_600,     # 16 files, 1020 insertions / 91 deletions
@@ -189,7 +189,7 @@ OBSERVED_RUNS: list[dict] = [
         "subagents": 0,
     },
     {
-        "label": "gitea_admin/pragent#7 (+ cost-model calibration + salvage fix)",
+        "label": "internal/hardening-PR (same PR, two commits later)",
         "date": "2026-08-18",
         "tier": "full",
         "diff_tokens": 21_000,     # same PR, two commits later

tests/pilot/test_ai_review.py

@@ -326,15 +326,18 @@ def test_inline_comment_body_severity_emoji_mapping():
         assert badge in inline_comment_body(f), f"{sev} → {badge}"
 
 
-def test_inline_comment_body_no_token_attribution():
+def test_inline_comment_body_with_token_attribution():
-    # Per spec: no per-comment 🪙 token attribution line.
+    # Operator wants per-comment attribution back: every inline comment shows
+    # the attributed output tokens + share of total. Hidden only when no
+    # attribution data was computed (legacy callers / ollama path without
+    # usage metering).
     f = {"severity": "high", "path": "a", "line": 1, "problem": "p",
          "fix": "f", "suggestion": "", "reference": "",
-         "_tok_attrib": 1234, "_tok_pct": 0.3}
+         "_tok_attrib": 1234, "_tok_pct": 0.30}
     body = inline_comment_body(f)
-    assert "🪙" not in body
+    assert "🪙 ~1234 tok" in body
-    assert "tok" not in body.lower().split("fix")[0]  # only in fix is OK
+    assert "30%" in body
-    assert "attributed" not in body
+    assert "attributed output" in body
 
 
 def test_summary_bullets_format():
@@ -673,15 +676,17 @@ def test_compute_attribution_noop_on_empty_or_zero_budget():
     assert "_tok_attrib" not in fs[0]
 
 
-def test_inline_comment_body_no_attribution_line():
+def test_inline_comment_body_with_attribution_line():
-    # Per spec: NO per-comment token attribution — that telemetry lives in the
+    # Operator wants per-comment attribution back: every inline comment shows
-    # collapsible block on the PR-level comment.
+    # the attributed output tokens + share of total. Hidden only when no
+    # attribution data was computed (legacy callers / ollama path without
+    # usage metering).
     f = {"severity": "high", "path": "a", "line": 1, "problem": "bad", "fix": "swap",
          "suggestion": "", "_tok_attrib": 180, "_tok_pct": 0.29}
     body = inline_comment_body(f)
-    assert "🪙" not in body
+    assert "🪙 ~180 tok" in body
-    assert "tok" not in body
+    assert "29%" in body
-    assert "attributed" not in body
+    assert "attributed output" in body
 
 
 def test_inline_comment_body_no_attribution_no_coin_line():

tests/pilot/test_cost_model.py

@@ -226,7 +226,9 @@ def test_observed_report_prices_every_model():
     text = cm.observed_report(["claude-opus-5", "gpt-5.6-luna"])
     assert "Claude Opus 5" in text
     assert "GPT-5.6 Luna" in text
-    assert "pragent#7" in text
+    # Labels are generic (no internal repo names) for commercialization.
+    assert "gitea_admin" not in text
+    assert "internal/hardening-PR" in text
 
 
 def test_model_is_within_an_order_of_magnitude_of_the_measurement():