zero-slop 2.12.15 → 2.12.18

package/README.md

@@ -14,7 +14,7 @@ Zero Slop is a free, open-source agent skill that finds and removes AI slop whil
 
 <p align="center">
   <a href="https://github.com/manavmishra/ZeroSlop/actions/workflows/validate.yml"><img alt="Validate" src="https://github.com/manavmishra/ZeroSlop/actions/workflows/validate.yml/badge.svg"></a>
-  <img alt="Version 2.12.15" src="https://img.shields.io/badge/version-2.12.15-72528F?color=C15732">
+  <img alt="Version 2.12.18" src="https://img.shields.io/badge/version-2.12.18-72528F?color=C15732">
   <a href="https://www.npmjs.com/package/zero-slop"><img alt="npm version" src="https://img.shields.io/npm/v/zero-slop?color=C15732"></a>
   <a href="https://www.npmjs.com/package/zero-slop"><img alt="npm downloads" src="https://img.shields.io/npm/dm/zero-slop?color=17634F"></a>
   <a href="https://github.com/manavmishra/ZeroSlop/stargazers"><img alt="GitHub stars" src="https://img.shields.io/github/stars/manavmishra/ZeroSlop?style=flat&color=C15732"></a>
@@ -176,7 +176,7 @@ python3 scripts/slopscore.py --batch drafts/ --gate 25
 The CLI sends a file to the hosted editor without changing the file on disk:
 
 ```sh
-npx --yes zero-slop@2.12.15 deslop draft.md --genre professional
+npx --yes zero-slop@2.12.18 deslop draft.md --genre professional
 ```
 
 Use `-` for stdin and `--json` for structured output. `--require-approved` prints the result but exits nonzero when review is needed. Requires Node.js 22+; offline `score` also needs Python 3. [CLI options and privacy](docs/cli.md).

package/SECURITY.md

@@ -17,6 +17,18 @@ The installed skill ships ten standard-library Python modules:
 | `scripts/safeio.py` | locks and atomic file replacement | only on behalf of the two writers above | none |
 | `scripts/version_check.py` | optional release check | none | one metadata-only GitHub API request |
 
+The Anthropic plugin folder additionally carries three host execution adapters:
+`mcp_stdio.py`, `mcp_local_tools.py`, and `mcp_workflow_tools.py`. They expose the
+existing local checks through a declared stdio MCP server; they do not run shell
+commands or accept caller-selected file paths. Learning writes still require
+affirmative consent, with separate consent for activation. The optional version
+tool uses the same metadata-only release check. These adapters and their routing
+overlay are absent from the portable skill, npm package, and OpenAI ZIP. The
+Anthropic folder retains the separate remote connector for explicitly requested
+hosted editing; local checks do not automatically invoke it.
+The declared Python entry point in a repository subfolder can be held for manual
+directory review. A declaration or local test does not establish scan approval.
+
 These local Python checks do not transmit the draft. Editing inside an assistant
 follows that assistant's privacy settings. The version checker sends only a GET
 for the latest public release tag, times out after 2.5 seconds, fails open, and can be

package/SKILL.md

@@ -2,7 +2,7 @@
 name: zero-slop
 license: MIT
 metadata:
-  version: "2.12.15"
+  version: "2.12.18"
   author: manavmishra
 description: Edit drafts into natural prose, inspect AI-sounding patterns, or review how a specified audience might respond passage by passage. Zero Slop runs inside the user's existing AI assistant with local tools that protect source details. Use for humanizing or de-slopping writing, polishing outward-facing prose, social drafts, final editorial checks, or an explicit simulated reader review. Preserve facts, voice and format; reader simulations are hypotheses, not human feedback.
 ---

package/docs/cli.md

@@ -3,7 +3,7 @@
 Use `deslop` for hosted editing with MCP parity. Use `score` for offline checks.
 
 ```sh
-npm install --global zero-slop@2.12.15
+npm install --global zero-slop@2.12.18
 zero-slop deslop draft.md --genre professional
 zero-slop deslop - --genre email < draft.txt
 zero-slop deslop draft.md --json --require-approved

package/gemini-extension.json

@@ -1,6 +1,6 @@
 {
   "name": "zero-slop",
-  "version": "2.12.15",
+  "version": "2.12.18",
   "description": "Find and fix canned AI writing while preserving the facts, voice, and format.",
   "mcpServers": {
     "zero-slop": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "zero-slop",
-  "version": "2.12.15",
+  "version": "2.12.18",
   "description": "An editing skill, offline writing checks, and a CLI for the hosted Zero Slop MCP pipeline. Find stock AI phrasing, rewrite, and check source details.",
   "bin": {
     "zero-slop": "bin/zero-slop.mjs"
@@ -56,6 +56,9 @@
     "!scripts/build_onepager_pdf.py",
     "!scripts/build_plugin.py",
     "!scripts/build_marketplace_package.py",
+    "!scripts/mcp_stdio.py",
+    "!scripts/mcp_local_tools.py",
+    "!scripts/mcp_workflow_tools.py",
     "!scripts/build_skill_zip.py",
     "!scripts/check_svg.py",
     "!scripts/check_distribution_manifests.py",

package/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
   "name": "zero-slop",
-  "version": "2.12.15",
+  "version": "2.12.18",
   "description": "Find and fix canned AI writing while preserving the facts, voice, and format.",
   "author": {
     "name": "Manav Mishra",

package/scripts/calibrate.py

@@ -95,8 +95,13 @@ def freqs(texts):
 
 
 def excess_weights(human_dir, ai_dir):
-    hc, hn = freqs(read_corpus(human_dir))
-    ac, an = freqs(read_corpus(ai_dir))
+    return excess_weights_from_texts(read_corpus(human_dir), read_corpus(ai_dir))
+
+
+def excess_weights_from_texts(human_texts, ai_texts):
+    """Same frequency comparison for supplied texts; no paths or persistence."""
+    hc, hn = freqs(human_texts)
+    ac, an = freqs(ai_texts)
     out = {}
     for term, n_ai in ac.items():
         if n_ai < MIN_OBS or len(term) < 4:

package/scripts/learn.py

@@ -656,8 +656,8 @@ def diff_spans(produced, shipped):
     return out
 
 
-def load_edit_feedback(path, produced, shipped, diffs):
-    """Load optional per-edit reason labels bound to an exact before/after pair."""
+def read_edit_feedback(path):
+    """Read a bounded feedback file once; semantic validation is separate."""
     source = Path(path)
     if not source.is_file():
         raise SystemExit(f"feedback is not a readable file: {source}")
@@ -667,6 +667,16 @@ def load_edit_feedback(path, produced, shipped, diffs):
         payload = json.loads(source.read_text())
     except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
         raise SystemExit(f"feedback must be readable UTF-8 JSON: {exc}") from exc
+    return payload
+
+
+def load_edit_feedback(path, produced, shipped, diffs):
+    """Load optional per-edit reason labels bound to an exact before/after pair."""
+    return validate_edit_feedback(read_edit_feedback(path), produced, shipped, diffs)
+
+
+def validate_edit_feedback(payload, produced, shipped, diffs):
+    """Validate source-bound edit labels without accepting a caller file path."""
     if (not isinstance(payload, dict)
             or set(payload) != {"schema", "source_sha256", "target_sha256", "edits"}
             or payload.get("schema") != 1
@@ -693,9 +703,28 @@ def load_edit_feedback(path, produced, shipped, diffs):
     return labels
 
 
-@state_locked(lambda *a, **k: OBS)
 def reflect(produced, shipped, doc_id=None, *, reason="unspecified", genre="general",
             feedback=None):
+    prod_text = required_text(produced, "produced draft")
+    ship_text = required_text(shipped, "shipped draft")
+    payload = None
+    if feedback:
+        # Keep the file interface and its existing bound for CLI callers.
+        payload = read_edit_feedback(feedback)
+    return reflect_text(prod_text, ship_text, reason=reason, genre=genre,
+                        feedback=payload, labels=(Path(produced).name, Path(shipped).name))
+
+
+@state_locked(lambda *a, **k: OBS)
+def reflect_text(prod_text, ship_text, *, reason="unspecified", genre="general",
+                 feedback=None, labels=("supplied draft", "supplied final text")):
+    """Record explicitly approved supplied edits using the existing safety gates.
+
+    No source files are written. Private evidence remains under ZERO_SLOP_HOME;
+    callers must obtain human opt-in before invoking this state-changing operation.
+    """
+    if not isinstance(prod_text, str) or not isinstance(ship_text, str):
+        raise ValueError("Reflection inputs must be text")
     reason = validate_label(reason, REASON_LABELS, "reason label")
     genre = validate_label(genre, GENRES, "genre")
     base = load(DATA / "patterns.json")
@@ -710,8 +739,6 @@ def reflect(produced, shipped, doc_id=None, *, reason="unspecified", genre="gene
     obs.setdefault("fix_observations", {})
 
     today = str(date.today())
-    prod_text = required_text(produced, "produced draft")
-    ship_text = required_text(shipped, "shipped draft")
     # A vote is one unique edit pair, regardless of filenames or caller-supplied
     # labels. Otherwise three copies of the same before/after text could cross
     # the recurrence threshold by changing only --doc-id. The optional label is
@@ -719,8 +746,8 @@ def reflect(produced, shipped, doc_id=None, *, reason="unspecified", genre="gene
     doc = hashlib.sha256(
         (prod_text + "\0" + ship_text).encode()).hexdigest()[:16]
     diffs = diff_spans(prod_text, ship_text)
-    edit_labels = (load_edit_feedback(feedback, prod_text, ship_text, diffs)
-                   if feedback else {})
+    edit_labels = (validate_edit_feedback(feedback, prod_text, ship_text, diffs)
+                   if feedback is not None else {})
     caught_words = {word for d in diffs if already_caught(d["span"], pats, lex)
                      for word in norm(d["span"]).split()}
 
@@ -818,7 +845,7 @@ def reflect(produced, shipped, doc_id=None, *, reason="unspecified", genre="gene
 
     write_json(OBS, obs, private=True)
 
-    print(f"reflect: {Path(produced).name} → {Path(shipped).name}\n")
+    print(f"reflect: {labels[0]} → {labels[1]}\n")
     print(f"  {agreed} edit(s) Zero Slop had already flagged — the writer agreed")
     print(f"  {skipped} content-specific cut(s) ignored (figures, proper nouns)")
     print(f"  {recorded} new phrase(s) saved for review")
@@ -1242,11 +1269,8 @@ def merge(path, apply_, cat, weight):
     return 0
 
 
-@state_locked(lambda *a, **k: LOCAL)
 def confirm(target):
     """Refresh evidence. Patterns that keep firing stay; the rest decay out."""
-    p = LOCAL
-    d = load_learned(p, "local")
     t = Path(target)
     if not t.exists():
         raise SystemExit(f"confirmation path does not exist: {t}")
@@ -1260,6 +1284,15 @@ def confirm(target):
         raise SystemExit(f"cannot read confirmation corpus at {t}: {exc}") from exc
     if not text.strip():
         raise SystemExit(f"confirmation corpus contains no text: {t}")
+    return confirm_text(text, sample_count=len(files))
+
+
+@state_locked(lambda *a, **k: LOCAL)
+def confirm_text(text, *, sample_count=1):
+    """Refresh private rules against supplied text, without caller paths."""
+    if not isinstance(text, str) or not text.strip():
+        raise ValueError("Confirmation input must contain text")
+    d = load_learned(LOCAL, "local")
     today, n = str(date.today()), 0
     for pat in d.get("patterns", []):
         try:
@@ -1270,9 +1303,9 @@ def confirm(target):
                 n += 1
         except re.error:
             continue
-    write_json(p, d, private=True)
+    write_json(LOCAL, d, private=True)
     print(f"confirmed {n}/{len(d.get('patterns', []))} learned pattern(s) "
-          f"against {len(files)} file(s)")
+          f"against {sample_count} sample(s)")
     return 0
 
 
@@ -1427,10 +1460,7 @@ def build_voice(name, sample_path):
     complete writing style. The resulting profile changes scoring only when
     the caller explicitly selects it with ``--voice NAME``.
     """
-    import slopscore
     name = safe_voice_name(name)
-    base = slopscore.load_patterns()
-    terms = list(base.get("lexicon", {})) + list(base.get("riders", {}))
     src = Path(sample_path)
     if not src.exists():
         raise SystemExit(f"voice sample path does not exist: {src}")
@@ -1444,6 +1474,18 @@ def build_voice(name, sample_path):
         raise SystemExit(f"cannot read voice sample at {src}: {exc}") from exc
     if not blob.strip():
         raise SystemExit(f"voice sample contains no text: {src}")
+    return build_voice_text(name, blob)
+
+
+def build_voice_text(name, text):
+    """Same exact-term profile from supplied text, never an arbitrary source path."""
+    import slopscore
+    name = safe_voice_name(name)
+    if not isinstance(text, str) or not text.strip():
+        raise ValueError("Voice sample must contain text")
+    base = slopscore.load_patterns()
+    terms = list(base.get("lexicon", {})) + list(base.get("riders", {}))
+    blob = text.lower()
     keep = sorted({term for term in terms
                    if re.search(r"\b" + re.escape(term.lower()) + r"\b", blob)})
     prof = {"_comment": f"Named scoring profile for {name}. Exact matches for "

package/scripts/slopscore.py

@@ -1490,6 +1490,11 @@ def load_adjudication(path, original):
         payload = json.loads(source.read_text(encoding="utf-8"))
     except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
         raise ValueError(f"cannot read adjudication file: {exc}") from exc
+    return validate_adjudication(payload, original)
+
+
+def validate_adjudication(payload, original):
+    """Validate the same source-bound ruling supplied as JSON, without file I/O."""
     if not isinstance(payload, dict):
         raise ValueError("adjudication file must be a JSON object")
     expected = {"schema", "original_sha256", "allow_dropped_figures"}