Local video editor for coding agents: 42 FFmpeg tools with a machine-readable contract and an MCP server, for Claude Code, Cursor and Codex. Cut, captions, reframe, loudness, platform exports. No cloud, no API keys.
Python agent-skillsai-agentsaudio-processingcaptionsclaude-codecodexcoding-agentcursorffmpegloudness-normalizationmcpmcp-servermodel-context-protocolpythonsubtitlestranscodingvideovideo-editingvideo-processingwhisper
fix: MCP default tools/list is the core 12, opt-in for the full 42 (#252)
Shogo Kajiyama8b1cd9fdfd7d
diff --git a/README.md b/README.mdindex 2161940..8429c4c 100644--- a/README.md+++ b/README.md@@ -41,7 +41,7 @@ npx ffmpeg-skillLeft half is the input, right half is what the command produced. **[All 53 before/after demos, with the exact command under each one →](docs/demos.md)** — all of it generated from synthetic footage by `python3 demos/build.py`, so you can rebuild every frame of it yourself.-`ffmpeg-skill` is an [Agent Skill](https://docs.anthropic.com/en/docs/agents-and-tools/agent-skills) for Claude Code, Cursor, Codex and any agent that reads `SKILL.md`. It teaches the agent a fixed workflow (probe → edit losslessly where possible → check → verify) and ships **42 tools** that do the actual work with `ffmpeg` / `ffprobe`: cut, join, silence removal, fit to duration and aspect, captions and karaoke, overlays and motion graphics, HDR → SDR and LUTs, audio clean-up and typed dynamics, sync with drift correction, multicam, loudness, delivery checks, whole-edit project rendering, batch folders. Every tool is also an MCP tool, and the whole set is described by a machine-readable contract.+`ffmpeg-skill` is an [Agent Skill](https://docs.anthropic.com/en/docs/agents-and-tools/agent-skills) for Claude Code, Cursor, Codex and any agent that reads `SKILL.md`. It teaches the agent a fixed workflow (probe → edit losslessly where possible → check → verify) and ships **42 tools** that do the actual work with `ffmpeg` / `ffprobe`: cut, join, silence removal, fit to duration and aspect, captions and karaoke, overlays and motion graphics, HDR → SDR and LUTs, audio clean-up and typed dynamics, sync with drift correction, multicam, loudness, delivery checks, whole-edit project rendering, batch folders. Every tool is also callable as an MCP tool (`tools/list` advertises a core 12 by default to keep client context small, with `FFMPEG_SKILL_MCP_FULL=1` listing all 42; every tool is reachable by name through `tools/call` either way), and the whole set is described by a machine-readable contract.If `ffmpeg` and `python3` are on your PATH, it works: offline, on footage you would rather not upload.@@ -341,9 +341,9 @@ Next to the tool list the document carries a top-level `deprecated` list (1.10):On Windows, `python3` is only on PATH if Python was installed from the Microsoft Store; a python.org install exposes `python` (or the `py` launcher) instead — if your MCP client reports the server failed to start, change `"command"` above to `"python"` (or the full path from `where python`).-`mcp/server.py` is a stdio JSON-RPC transport with no tool table of its own. `tools/list` is derived from the contract at start-up: the same 42 names, the same order, and `inputSchema` translated from each tool's `input_schema`. `tools/call` maps structured arguments to argv and runs the named script; a raw `argv` form is accepted for compatibility and marked non-canonical. `python3 mcp/server.py --list` prints the tools; `--call probe '{"inputs": ["a.mp4"]}'` runs one from the shell.+`mcp/server.py` is a stdio JSON-RPC transport with no tool table of its own. `tools/list` is derived from the contract at start-up, in contract order, with `inputSchema` translated from each tool's `input_schema`. By default it lists only the core 12 (`render`, `look`, `caption`, `export`, `check`, `fit`, `cut`, `audio`, `loudness`, `graphics`, `silence`, `probe`) so a client doesn't pay context for 30 schemas it rarely calls directly; set `FFMPEG_SKILL_MCP_FULL=1` to list all 42. Every tool, listed or not, is callable through `tools/call`, which maps structured arguments to argv and runs the named script; a raw `argv` form is accepted for compatibility and marked non-canonical. `python3 mcp/server.py --list` prints the tools; `--call probe '{"inputs": ["a.mp4"]}'` runs one from the shell.-`FFMPEG_SKILL_MCP_LEAN=1` in the server's environment drops `json` and `progress` from every `inputSchema`: they are transport flags the server sets itself, not tool arguments, and 2.0 drops them unconditionally. It is opt-in, so the default `tools/list` stays byte-identical to the CLI surface the contract promises.+`FFMPEG_SKILL_MCP_LEAN=1` in the server's environment drops `json` and `progress` from every `inputSchema`: they are transport flags the server sets itself, not tool arguments, and 2.0 drops them unconditionally. It is opt-in, independent of `FFMPEG_SKILL_MCP_FULL`, and the default `tools/list` stays byte-identical (aside from the core-12 filter) to the CLI surface the contract promises.### Capability detectiondiff --git a/docs/contract.md b/docs/contract.mdindex 5d64d62..203e8f2 100644--- a/docs/contract.md+++ b/docs/contract.md@@ -578,6 +578,18 @@ for good (see "What 2.0 changes"). The flag is opt-in and changes nothing else:pins -- descriptions may change between releases (the `--crf` deprecation mark did) -- so a leanclient and a default client see the same tools with the same names.+### Default `tools/list` surface: core 12, opt-in for all 42++By default `tools/list` returns only the core dozen in `_contract.MCP_CORE_TOOLS` (`render`,+`look`, `caption`, `export`, `check`, `fit`, `cut`, `audio`, `loudness`, `graphics`, `silence`,+`probe`) rather than all 42, so a session doesn't pay context for 30 schemas it is unlikely to ever call+directly -- the set is the tools eval iterations 17-20's ground-truth `expect` lists actually name+most often across the corpus in `evals/agent_prompts*.json`. Every tool -- including the other 30+-- is still callable by name through `tools/call` regardless of what `tools/list` advertised; the+contract (`ffmpeg-skill contract --json`) still describes all 42 unconditionally. Set+`FFMPEG_SKILL_MCP_FULL=1` (anything but "" or `0`) in the server's environment to make `tools/list`+return all 42, as every version before 1.18.3 did.+## Consuming the contract from an agentA planning agent (for example video-production-agent's SkillRegistry) can:diff --git a/docs/design-decisions.md b/docs/design-decisions.mdindex 8332fc2..73ceb9b 100644--- a/docs/design-decisions.md+++ b/docs/design-decisions.md@@ -201,9 +201,11 @@ decided. Recorded here so none of it is re-proposed from scratch.so the rule ("these are sentences, not labels") has an example next to it.- **P1-6 — eval prompt `ml2` tested a refusal, not the mux.** Accepted: `ml2` now ships a GermanSRT and a video and expects `caption|render`.-- **P1-7 — the 42-tool MCP `tools/list` is paid for by every session.** Accepted, and moved- **ahead of the optical-flow work into 1.18.0**: the default listing becomes the core 12, the- rest stay reachable through the contract, which still describes all 42.+- **P1-7 — the 42-tool MCP `tools/list` is paid for by every session.** Accepted, and **shipped**:+ `tools/list` defaults to the core 12 (`render`, `look`, `caption`, `export`, `check`, `fit`,+ `cut`, `audio`, `loudness`, `graphics`, `silence`, `probe`), chosen from eval iterations 17-20's+ `expect` frequency, not taste. `FFMPEG_SKILL_MCP_FULL=1` lists all 42; the other 30 stay+ reachable by name through `tools/call` either way, and the contract still describes all 42.- **P1-8 — at least one routing run on a non-Claude model.** Accepted as an `evals/run.py` task:the harness reads transcripts and files and is not Claude-specific; the routing/refusal set iswhat a Cursor or Codex run would publish.@@ -253,9 +255,10 @@ Each item is recorded here so it is not re-proposed from scratch.1.11.1 reworded step 0 accordingly. The answer is not to make `references/` more attractive butto make the table better: the **intent-clustered request table is planned for 1.20.0**.- **The MCP catalogue costs a client context on every session.** A 42-tool `tools/list` is paid- for by every session, including the ones that call two tools. Decided for **1.19.0**: the- default listing becomes the core 12, the other 30 are reachable lazily through the contract.- The contract still describes all 42 — the surface does not shrink, only the default listing.+ for by every session, including the ones that call two tools. Decided and shipped in **1.18.3**:+ the default listing becomes the core 12, the other 30 are reachable lazily through `tools/call`+ or `FFMPEG_SKILL_MCP_FULL=1`. The contract still describes all 42 — the surface does not shrink,+ only the default listing.- **The roadmap read as though planned work had shipped.** Fixed: `docs/roadmap.md` now marksevery version shipped + evaluated (naming the iteration), shipped with eval pending, orplanned, and the reconciliation is against `CHANGELOG.md` and `evals/results/`.diff --git a/mcp/server.py b/mcp/server.pyindex 59b7d71..eac7311 100755--- a/mcp/server.py+++ b/mcp/server.py@@ -1,10 +1,13 @@#!/usr/bin/env python3"""ffmpeg-skill as an MCP server (stdio, JSON-RPC 2.0) — standard library only.-Every public script in ../scripts becomes a tool. Tool names, order, inputSchema and the-structured-argument mapping all come from the contract (scripts/_contract.py); this file-is only the transport. Arguments are passed as a flat object (keys = argparse dests) or,-for CLI compatibility, as an argv list. Results are the script's --json output.+Every public script in ../scripts becomes a callable tool, but tools/list defaults to+advertising only the core 12 (_contract.MCP_CORE_TOOLS) to keep the schema dump small;+set FFMPEG_SKILL_MCP_FULL=1 to list all 42. Every tool is still callable by name via+tools/call either way. Tool names, order, inputSchema and the structured-argument mapping+all come from the contract (scripts/_contract.py); this file is only the transport.+Arguments are passed as a flat object (keys = argparse dests) or, for CLI compatibility,+as an argv list. Results are the script's --json output.Run:python3 mcp/server.py # stdio transport@@ -126,8 +129,18 @@ def call_tool(name: str, args: Dict[str, Any]) -> Dict[str, Any]:def tool_list() -> List[Dict[str, Any]]:- """tools/list: one entry per ToolSpec, in the contract's order, inputSchema derived from ToolSpec.input_schema."""- return [_contract.mcp_tool(spec) for spec in specs().values()]+ """tools/list: one entry per ToolSpec, in the contract's order, inputSchema derived from ToolSpec.input_schema.++ Default surface is the core 12 (_contract.MCP_CORE_TOOLS, roadmap P1-7): the other 30 stay+ reachable by name through tools/call regardless. FFMPEG_SKILL_MCP_FULL=1 (anything but "" or+ "0") lists all 42, unchanged from pre-1.18.3 behaviour.+ """+ all_specs = specs()+ if os.environ.get("FFMPEG_SKILL_MCP_FULL", "") not in ("", "0"):+ names = all_specs.keys()+ else:+ names = (n for n in all_specs if n in _contract.MCP_CORE_TOOLS)+ return [_contract.mcp_tool(all_specs[name]) for name in names]def handle(req: Dict[str, Any]) -> Dict[str, Any]:diff --git a/scripts/_contract.py b/scripts/_contract.pyindex 2a6bb52..8f8076a 100644--- a/scripts/_contract.py+++ b/scripts/_contract.py@@ -1075,6 +1075,15 @@ MCP_JSON_EXEMPT = ("look", "probe")# itself, not tool arguments. Off by default so tools/list stays byte-identical to the CLI surface# the contract promises; 2.0 drops them unconditionally.MCP_LEAN_DROP = ("json", "progress")+# default MCP tools/list surface (roadmap P1-7, shipped 1.18.3): the 42-tool schema dump costs+# every MCP session context whether it needs it or not. These 12 are the tools eval iterations+# 17-20's ground-truth `expect` lists actually name most often across the 118 agent prompts in+# evals/agent_prompts*.json (render/look/caption/export/check each >=10 hits; fit/cut/audio/+# loudness/graphics/silence/probe each >=6) -- not a hand-picked "starter" set. `tools/list`+# returns only these unless FFMPEG_SKILL_MCP_FULL=1 (anything but "" or "0") asks for all 42; the+# other 30 stay reachable by name through tools/call either way, and `contract --json` always+# describes all 42.+MCP_CORE_TOOLS = ("render", "look", "caption", "export", "check", "fit", "cut", "audio", "loudness", "graphics", "silence", "probe")MCP_STRUCTURED_NOTE = ("Structured arguments: keys are the input_schema property names (argparse dests), positionals ""are passed by name, output -> -o. Or argv: the raw CLI list (non-canonical; all other keys are then ignored). ""Media paths must be absolute.")diff --git a/tests/test_contract.py b/tests/test_contract.pyindex febe074..9b6cb78 100644--- a/tests/test_contract.py+++ b/tests/test_contract.py@@ -801,19 +801,44 @@ class ContractTests(unittest.TestCase):# ------------------------------------------------------------------ consistency: MCP and installerdef test_mcp_tools_match_contract(self):- mcp_names = [t["name"] for t in mcp_server.tool_list()]- self.assertEqual(sorted(mcp_names), sorted(self.tools), "MCP tools/list == contract tools")- listed = [l.split()[0] for l in sh(sys.executable, ROOT / "mcp" / "server.py", "--list").stdout.splitlines() if l.strip()]- self.assertEqual(sorted(listed), sorted(self.tools))+ """FFMPEG_SKILL_MCP_FULL=1 is the escape hatch back to the pre-1.18.3 behaviour: every+ contract tool listed. This test proves that view is still byte-for-byte the full set."""+ saved = os.environ.get("FFMPEG_SKILL_MCP_FULL")+ os.environ["FFMPEG_SKILL_MCP_FULL"] = "1"+ try:+ mcp_names = [t["name"] for t in mcp_server.tool_list()]+ self.assertEqual(sorted(mcp_names), sorted(self.tools), "MCP tools/list (full) == contract tools")+ listed = [l.split()[0] for l in sh(sys.executable, ROOT / "mcp" / "server.py", "--list", env={**os.environ}).stdout.splitlines() if l.strip()]+ self.assertEqual(sorted(listed), sorted(self.tools))+ finally:+ os.environ.pop("FFMPEG_SKILL_MCP_FULL", None)+ if saved is not None:+ os.environ["FFMPEG_SKILL_MCP_FULL"] = savedfor name, spec in self.tools.items():self.assertEqual(spec["mcp"]["tool"], name)self.assertEqual(spec["mcp"]["positional"], mcp_server.POSITIONAL.get(name, ["input"]) if spec["mcp"]["positional"] else spec["mcp"]["positional"])- # a real JSON-RPC round trip- req = json.dumps({"jsonrpc": "2.0", "id": 1, "method": "tools/list"}) + "\n"- proc = subprocess.run([sys.executable, str(ROOT / "mcp" / "server.py")], input=req, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)- resp = json.loads(proc.stdout.strip().splitlines()[0])+ # a real JSON-RPC round trip, full surface+ resp = self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}], full=True)[0]self.assertEqual(sorted(t["name"] for t in resp["result"]["tools"]), sorted(self.tools))+ def test_mcp_tools_list_defaults_to_the_core_12(self):+ """roadmap P1-7 (docs/design-decisions.md): tools/list defaults to _contract.MCP_CORE_TOOLS,+ in contract order, not all 42 -- and every one of those 12 is a real contract tool. The+ other 30 stay reachable through tools/call (proved by test_mcp_schema_drift_follows_the_scripts+ and test_mcp_server_error_paths_raw_argv_and_call_helper calling non-core tools directly)."""+ self.assertEqual(len(_contract.MCP_CORE_TOOLS), 12)+ self.assertTrue(set(_contract.MCP_CORE_TOOLS).issubset(self.tools), "every core tool is a real contract tool")+ expected_order = [n for n in self.tools if n in _contract.MCP_CORE_TOOLS]+ mcp_names = [t["name"] for t in mcp_server.tool_list()]+ self.assertEqual(mcp_names, expected_order, "default MCP tools/list is the core 12, in contract order")+ resp = self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}])[0]+ self.assertEqual([t["name"] for t in resp["result"]["tools"]], expected_order, "JSON-RPC round trip matches")+ for flag in ("0", ""):+ resp = self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}], full=False)[0]+ self.assertEqual([t["name"] for t in resp["result"]["tools"]], expected_order, f"FFMPEG_SKILL_MCP_FULL={flag!r} still core 12")+ listed = [l.split()[0] for l in sh(sys.executable, ROOT / "mcp" / "server.py", "--list").stdout.splitlines() if l.strip()]+ self.assertEqual(listed, expected_order)+def test_mcp_tool_surface_matches_the_frozen_1x_snapshot(self):"""docs/contract.md, "Stability guarantee (1.x)": within 1.x no tool is removed or renamedand no argument is removed, renamed or made newly required. The MCP surface is *derived*@@ -823,12 +848,21 @@ class ContractTests(unittest.TestCase):Adding a tool or an optional argument is allowed and must be reflected by regenerating thesnapshot (UPDATE_MCP_SNAPSHOT=1 python3 tests/test_contract.py); removing or renaminganything, or making an argument required, is a breaking change and belongs behind the- deprecation policy and a major bump, not behind a regenerated fixture."""+ deprecation policy and a major bump, not behind a regenerated fixture. The snapshot covers+ all 42 tools, so this reads tool_list() under FFMPEG_SKILL_MCP_FULL=1 -- the default's+ core-12 filter is a separate, tested concern (test_mcp_tools_list_defaults_to_the_core_12)."""snapshot_path = ROOT / "tests" / "fixtures" / "mcp_tools.json"- live = [{"name": t["name"],- "properties": sorted(t["inputSchema"].get("properties", {}).keys()),- "required": sorted(t["inputSchema"].get("required", []))}- for t in mcp_server.tool_list()]+ saved = os.environ.get("FFMPEG_SKILL_MCP_FULL")+ os.environ["FFMPEG_SKILL_MCP_FULL"] = "1"+ try:+ live = [{"name": t["name"],+ "properties": sorted(t["inputSchema"].get("properties", {}).keys()),+ "required": sorted(t["inputSchema"].get("required", []))}+ for t in mcp_server.tool_list()]+ finally:+ os.environ.pop("FFMPEG_SKILL_MCP_FULL", None)+ if saved is not None:+ os.environ["FFMPEG_SKILL_MCP_FULL"] = savedif os.environ.get("UPDATE_MCP_SNAPSHOT"):snapshot_path.write_text(json.dumps(live, indent=1) + "\n", encoding="utf-8")frozen = json.loads(snapshot_path.read_text(encoding="utf-8"))@@ -892,9 +926,14 @@ class ContractTests(unittest.TestCase):self.assertIn(surface, wheres)# ------------------------------------------------------------------ MCP inputSchema derived from the contract- def _rpc(self, requests, root=ROOT):+ def _rpc(self, requests, root=ROOT, full=False):text = "".join(json.dumps(r) + "\n" for r in requests)- proc = subprocess.run([sys.executable, str(Path(root) / "mcp" / "server.py")], input=text, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)+ env = dict(os.environ)+ if full:+ env["FFMPEG_SKILL_MCP_FULL"] = "1"+ else:+ env.pop("FFMPEG_SKILL_MCP_FULL", None)+ proc = subprocess.run([sys.executable, str(Path(root) / "mcp" / "server.py")], input=text, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, env=env)self.assertEqual(proc.returncode, 0, proc.stderr)return [json.loads(line) for line in proc.stdout.strip().splitlines()]@@ -903,7 +942,7 @@ class ContractTests(unittest.TestCase):return json.dumps(obj, sort_keys=True, separators=(",", ":"))def test_mcp_input_schema_equals_translated_contract_schema(self):- listed = self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}])[0]["result"]["tools"]+ listed = self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}], full=True)[0]["result"]["tools"]self.assertEqual([t["name"] for t in listed], [t["name"] for t in self.contract["tools"]], "MCP order == contract order")for entry in listed:spec = self.tools[entry["name"]]@@ -949,7 +988,7 @@ class ContractTests(unittest.TestCase):(root / "scripts" / "cut.py").unlink()# a new public script without metadata is reported, not silently dropped or guessed(root / "scripts" / "zzztool.py").write_text("#!/usr/bin/env python3\nimport argparse\ndef main():\n argparse.ArgumentParser().parse_args()\nif __name__ == '__main__':\n main()\n", encoding="utf-8")- err = self._rpc([{"jsonrpc": "2.0", "id": 0, "method": "tools/list"}], root=root)[0]+ err = self._rpc([{"jsonrpc": "2.0", "id": 0, "method": "tools/list"}], root=root, full=True)[0]self.assertIn("no TOOL_META entry", err["error"]["message"])contract_py = (root / "scripts" / "_contract.py").read_text(encoding="utf-8")contract_py = contract_py.replace('TOOL_META: Dict[str, Dict[str, Any]] = {', 'TOOL_META: Dict[str, Dict[str, Any]] = {\n "zzztool": dict(role="analysis", inputs=["x"], outputs=["y"], required=["ffprobe"], optional=[], video_required=False, audio_only=True, visual=False, verify=[], produces_artifact=False, idempotency="bit_exact", deterministic=True),', 1)@@ -964,13 +1003,22 @@ class ContractTests(unittest.TestCase):fit = (root / "scripts" / "fit.py").read_text(encoding="utf-8")fit = fit.replace(' add_common(ap)', ' ap.add_argument("--drift-flag", action="store_true", help="added for the drift test")\n add_common(ap)', 1)(root / "scripts" / "fit.py").write_text(fit, encoding="utf-8")- tools = {t["name"]: t for t in self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}], root=root)[0]["result"]["tools"]}+ tools = {t["name"]: t for t in self._rpc([{"jsonrpc": "2.0", "id": 1, "method": "tools/list"}], root=root, full=True)[0]["result"]["tools"]}self.assertNotIn("cut", tools, "removed script still exposed")self.assertIn("zzztool", tools, "new public script not exposed")self.assertEqual(tools["zzztool"]["inputSchema"]["properties"]["knob"], {"type": "integer", "description": "a knob", "enum": [1, 2, 3], "default": 3})self.assertIn("drift_flag", tools["fit"]["inputSchema"]["properties"], "parser change not reflected")self.assertEqual(list(tools), sorted(tools), "order stays sorted after changes")- # the temporary tool also runs through the derived mapping+ # the same drift, seen through the default core-12 view: zzztool never joins it (it+ # isn't in MCP_CORE_TOOLS), the removed "cut" stays gone, and a core tool's parser+ # change ("fit") still reaches it -- both views are derived from the same live scripts.+ core = {t["name"]: t for t in self._rpc([{"jsonrpc": "2.0", "id": 4, "method": "tools/list"}], root=root)[0]["result"]["tools"]}+ self.assertNotIn("cut", core, "removed script still exposed in the core-12 view")+ self.assertNotIn("zzztool", core, "a non-core new script must not appear in the default view")+ self.assertIn("drift_flag", core["fit"]["inputSchema"]["properties"], "parser change not reflected in the core-12 view")+ self.assertEqual(set(core) - {"cut"}, set(_contract.MCP_CORE_TOOLS) - {"cut"}, "core-12 view is exactly MCP_CORE_TOOLS minus the removed script")+ # the temporary tool also runs through the derived mapping (tools/call reaches it even+ # though it is outside the default tools/list view)call = self._rpc([{"jsonrpc": "2.0", "id": 2, "method": "tools/call", "params": {"name": "zzztool", "arguments": {"input": "x", "knob": 2}}}], root=root)[0]self.assertEqual(call["result"]["structuredContent"]["knob"], 2)unknown = self._rpc([{"jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": {"name": "_contract", "arguments": {}}}], root=root)[0]diff --git a/tests/test_orchestration.py b/tests/test_orchestration.pyindex e26e5ac..6cbf1cf 100644--- a/tests/test_orchestration.py+++ b/tests/test_orchestration.py@@ -843,7 +843,9 @@ class OrchestrationTests(MediaFixtures):def test_brand_defaults_apply_to_caption_and_logo(self):brand = OUT / "brand.json"if not brand.exists():- self.test_graphics_templates()+ brand.write_text(json.dumps({"font": "DejaVu Sans", "colors": {"primary": "FF6A00", "text": "FFFFFF", "background": "0B1D2A"},+ "logo": "logo.png", "logo_position": "top-right", "logo_scale": 140, "safe_margin": 40,+ "caption": {"size": 28, "position": "bottom", "animate": "pop", "karaoke": True, "bold": True}}), encoding="utf-8")ass = OUT / "brand.ass"script("caption.py", self.src, "--text", self.cues, "--brand", brand, "--write-ass", ass, "--fast", "-o", OUT / "brand_cap.mp4")style = [l for l in ass.read_text(encoding="utf-8-sig").splitlines() if l.startswith("Style:")][0]@@ -868,7 +870,11 @@ class OrchestrationTests(MediaFixtures):{"jsonrpc": "2.0", "id": 6, "method": "tools/call", "params": {"name": "nope", "arguments": {}}},{"jsonrpc": "2.0", "id": 7, "method": "bogus/method"},]- proc = subprocess.run([sys.executable, str(server)], input="\n".join(json.dumps(r) for r in reqs) + "\n", stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)+ # FFMPEG_SKILL_MCP_FULL=1: this test is about the stdio wiring (initialize / tools/list /+ # tools/call / unknown name / unknown method), not the core-12-by-default filter, which+ # has its own tests in tests/test_contract.py.+ env = {**os.environ, "FFMPEG_SKILL_MCP_FULL": "1"}+ proc = subprocess.run([sys.executable, str(server)], input="\n".join(json.dumps(r) for r in reqs) + "\n", stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True, env=env)self.assertEqual(proc.returncode, 0, proc.stderr)resp = {r["id"]: r for r in (json.loads(l) for l in proc.stdout.splitlines() if l.strip())}self.assertEqual(resp[1]["result"]["serverInfo"]["name"], "ffmpeg-skill")