Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add pandazki/pneuma-skills --skill skillgit clone --depth 1 https://github.com/pandazki/pneuma-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/pandazki/pneuma-skills/skill)<a href="https://agentmods.dev/skills/pandazki/pneuma-skills/skill"><img src="https://agentmods.dev/badge/skills/pandazki/pneuma-skills/skill/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/pandazki/pneuma-skills/skill"><img src="https://agentmods.dev/badge/skills/pandazki/pneuma-skills/skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Anti-Refusal · line 96 Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.Fix: Remove instructions that suppress warnings, disclaimers, or ethical commentary. Let the agent surface safety-relevant caveats to the user.
- high Anti-Refusal · line 550 Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.Fix: Remove instructions that suppress warnings, disclaimers, or ethical commentary. Let the agent surface safety-relevant caveats to the user.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00130 | $0.09690 |
| Opus 5.5 | $0.00052 | $0.03876 |
| Sonnet 5.5 | $0.00026 | $0.01938 |
| Haiku 4.5 | $0.00013 | $0.00969 |
Grade A, and why
pneuma-backlot scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 14d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 439 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Pneuma Backlot Skill
Scene
You are the crew of a very small studio and the creator is its director. One short film is made end to end: an idea becomes a screenplay; that becomes a cast with designed faces, a look and places described in words; those become a shot list with every beat's picture written down; the shots become takes a video model paints; the takes become a cut. Eight stages, in order, each leaving a file the next is built from, never from your memory.
The previz department is the heart of this studio and it works in one order: design the picture in words, block it, then buy it. An untextured Blender animation — a greybox — fixes the space, who is where and when, what the props do and what the camera does, and a video model paints that clip. A video model is a brilliant painter and a fine actor, and an unreliable set designer and camera operator: asked in words for "walks in, stops, touches the panel, then the device lights up", it invents the room, the timing and the camera. So in a blocked shot space, blocking, prop events and camera live in the greybox; body action, look and tempo live in the prompt, carried forward from the plan. Greybox people are pawns, not puppets: a body-sized volume with a head and a front.
The block is a tool, not the film (rule 12): spend it where geography or a camera move is the hard thing, and shoot the fight and charm beats free. Where a shot IS blocked the greybox is the only picture of layout, behaviour and camera, and every other picture is an appearance — each still tried beside it (a storyboard, a key frame, the set concept, the previous shot's out-frame) brought a composition that fought the block and the model averaged the two. So a blocked take carries the greybox, the sheets and one style key frame, each told what it is not for, and a join travels as words.
Viewer contract
One project is a top-level directory (a content set) holding
backlot.json, idea.md, screenplay.md, bible/, shots/<id>/…, sound/
and cut/. Across the top of the viewer is the stage rail: the eight
stages with a status pill (empty · draft · approved · changed, the last
in the warning colour) and what that stage cost. Under it, a body per stage:
What ships with it
32 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/bible.md 14 KB
- references/camera.md 15 KB
- references/greybox.md 30 KB
- references/prompting.md 36 KB
- references/recreate.md 6.3 KB
- references/screenplay.md 11 KB
- references/scripts.md 25 KB
- references/shot-plan.md 22 KB
- references/sound-and-cut.md 14 KB
- references/video-generation.md 17 KB
- scripts/backlot.mjs 99 KB runs code
- scripts/blender/previz_kit.py 107 KB runs code
- scripts/cost.d.mts 1.5 KB
- scripts/cost.mjs 8.4 KB runs code
- scripts/media.d.mts 2.4 KB
- scripts/media.mjs 12 KB runs code
- scripts/previz.mjs 231 KB runs code
- scripts/prices.d.mts 1.3 KB
- scripts/prices.mjs 4.2 KB runs code
- scripts/project.mjs 11 KB runs code
- scripts/scene-starter/examples/courtyard.py 9.1 KB runs code
- scripts/scene-starter/examples/door_pull.py 3.5 KB runs code
- scripts/scene-starter/examples/duel_collage.py 4.6 KB runs code
- scripts/scene-starter/examples/duel_crane.py 2.9 KB runs code
- scripts/scene-starter/examples/duel_dolly_zoom.py 3.9 KB runs code
- scripts/scene-starter/examples/duel_orbit.py 2.5 KB runs code
- scripts/scene-starter/examples/lab_walk.py 3.4 KB runs code
- scripts/scene-starter/scene.py 3.5 KB runs code
- scripts/shot.d.mts 15 KB
- scripts/shot.mjs 58 KB runs code
- scripts/stage-state.d.mts 1.8 KB
- scripts/stage-state.mjs 14 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 14d ago Changed · +5 lines bb7c8313f786
- 1mo ago Changed · +23 lines 960b514508db
- 1mo ago Changed · -173 lines · -20 tokens per session 00477a2d100a
- 1mo ago Changed · +150 tokens per session 94bd03f380ea
- 1mo ago First seen · 584 lines · 0 tokens per session scan A b1d11c3129d0
pneuma-backlot is a skill published in the GitHub repository pandazki/pneuma-skills (161 stars, last pushed today), licensed MIT. It adds 130 tokens to every session and 9,690 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
learning-buddy
An AI coding learning partner using Socratic questioning and the 4Ds methodology — never gives away direct answers. Trigger when the user wants to understand, not just solve: phrases like "help me understand", "walk me through", "I don't just want the fix", "I want to learn why", "don't write the code for me", or "be…
omh-codebase-uml
This is an OMH codebase-uml workflow skill, projected for Agent Skills hosts (Claude Code, Codex, Cursor, opencode, OpenClaw, pi).
Teach From Sources
Read the links or sources the user gives, in the workspace browser, and teach the topic from them — grounded, ordered, and honest about what the sources do not say.
ralph
Launcher for the Ralph-style queue drain script - emits the right node ralph-loop.mjs invocation based on the user's filter and bounds. The actual loop runs as a Node script that opens one server run and spawns one claude -p --worktree per iteration under its own seat; this skill is the configurator, not the loop. Use…
session-retrospective
Analyzes the current implementation run - evaluates schema effectiveness, delegation alignment, note quality, and plan-to-execution fit. Captures cross-session trends and proposes improvements when patterns repeat. Use after implementation runs, or when user says 'retrospective', 'session review', 'what did we learn'…
adopt-project-scope
Migrates an already-populated, unscoped Task Orchestrator database in place to the project-scoping convention — creates a project anchor root, re-parents existing work trees under it, and writes rootId back to config.yaml. Use when a user says: adopt project scope, migrate this database to project scoping, make this…