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 Dongwu259/photo_s --skill photo-sgit clone --depth 1 https://github.com/Dongwu259/photo_sWrote 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/dongwu259/photo_s/photo-s)<a href="https://agentmods.dev/skills/dongwu259/photo_s/photo-s"><img src="https://agentmods.dev/badge/skills/dongwu259/photo_s/photo-s/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/dongwu259/photo_s/photo-s"><img src="https://agentmods.dev/badge/skills/dongwu259/photo_s/photo-s.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00443 | $0.04004 |
| Opus 5 | $0.00221 | $0.02002 |
| Sonnet 5 | $0.00089 | $0.00801 |
| Haiku 4.5 | $0.00044 | $0.00400 |
Grade A, and why
photo-s 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 yesterday.
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 — 170 lines — stays where its author put it; the contents beside it link to each section on GitHub.
PhotoS — Batch Photo Toolbox
Positioning: CLI for AI agents, GUI for humans. All heavy lifting is exposed as stable CLI commands with machine-readable JSON output. Full contract:
docs/AGENT_API.md(in this repo).
Install
pip install photo-s-tools # core: Pillow + rawpy (RAW decode)
pip install "photo-s-tools[enhance]" # + opencv: denoise / straighten / HDR / face blur
pip install "photo-s-tools[mcp]" # + MCP server (Python 3.10+)
- Binary:
photo-s· Python package:photo_s(importable in-process) · PyPI name:photo-s-tools· brand: PhotoS. - Works on macOS / Linux / Windows. Python 3.9+.
Golden rules (read first)
- Always pass
--jsonfor automation. Every command returns one JSON document on stdout with a top-levelschema_versionkey (currently1); human-readable text goes to stderr when--jsonis set. JSON keys are always English. - Never delete or overwrite inputs implicitly.
batch/compresswrite outputs beside inputs (or into-o DIR). Deletion-style commands (dedup keep-sharpest,--remove-original) require explicit flags. - Prefer
--dry-runfirst for anything that moves or deletes files (select,dedup) — zero writes, same JSON shape withwould_*actions. - Pin language when text matters:
--language en|zh|auto(defaultauto= system detection). Useenfor scripted/CI use. - Exit codes:
0= success,1= failure (bad args, missing deps). Per-file errors are reported inresults[].status/error, not by aborting the batch. - Optional opencv features (
denoise,auto_straighten,hdr,blurfaces) fail per-file with a clear "install photo-s-tools[enhance]" message when opencv is missing. Checkphoto-s info --json→optional_featuresfirst.
Subcommand → task map
| Task | Command | Notes |
|---|---|---|
| Full pipeline batch (compress/convert/resize/tone/crop/watermark/…) | photo-s batch PATHS... [-o DIR] [-q 85] [--resize 1920x1080] [--crop-ratio 16:9] [--strip-gps] [--preset NAME] [--blur-faces MODE] [--denoise] [--ev -1] |
MODE = blur or pixelate; see photo-s batch --help |
| Single-file convert | photo-s convert IN [-o OUT] |
|
| RAW → JPEG/TIFF batch | photo-s batch 'RAW/*.ARW' --format jpeg |
rawpy built-in, no extra dep |
| EXIF read / write / filter | photo-s exif PATHS... [--rating N] [--keywords ...] [--gps "lat,lon"] [--make --model] [--from-csv meta.csv] |
--rating-min N filters |
| Deduplicate | photo-s dedup PATHS... [--action keep-sharpest] [--dry-run] |
perceptual hash; actions: report/move/delete/keep-sharpest |
| Cull by exposure/sharpness | photo-s cull PATHS... [--overexposed-max ...] [--sharpness-min ...] |
threshold pass rejects |
| Quality-score ranking | photo-s cull PATHS... --score [--json] |
weighted 0-100 (exposure/contrast/sharpness/saturation) - ranks, never rejects |
| Burst keep-best | photo-s cull PATHS... --burst [--gap 2] --list |
clusters frames by EXIF time, keeps the top score per burst |
| Rank & move by rating | photo-s select PATHS... [--keep-min 4] [--reject-max 2] [--selects-dir DIR] [--rejects-dir DIR] [--copy] [--dry-run] |
reads EXIF rating |
| HDR merge (exposure brackets) | photo-s hdr IMG1 IMG2 IMG3... -o out.jpg [--align] |
needs [enhance] |
| Face blur / pixelate | photo-s blurfaces PATHS... [--mode pixelate] [--margin 20] [-o DIR] |
privacy; needs [enhance] |
| Contact sheet | photo-s contact-sheet PATHS... -o sheet.jpg [--cols 4] [--caption] |
|
| HTML gallery | photo-s gallery PATHS... -o OUTDIR [--title ...] |
|
| Batch rename | photo-s rename PATHS... --pattern '{seq}' |
live preview; vars: {year} {month} {day} {date} {time} {camera} {make} {iso} {focal} {seq} |
| Corrupt-file scan | photo-s check PATHS... [--json] |
integrity report |
| SHA-256 manifest | photo-s hash PATHS... -o manifest.csv [--verify manifest.csv] |
generate / verify |
| Watch a folder | photo-s watch DIR [-o DIR] [--recursive] |
auto-process new files |
| Presets | photo-s preset save NAME --quality 80 ... then photo-s batch ... --preset NAME |
saves full option set |
| Speed benchmark | photo-s bench --dir DIR -j 1,2,4,8 [--evaluate] |
temp output, source untouched |
| LR-style grading | photo-s batch PATHS... --curves "0,0;128,140;255,255" [--levels --vibrance --hsl --point-color --clarity --dehaze] |
compact-string params, see batch --help |
| Local masks + adjustments | photo-s batch PATHS... --masks "sky:linear:0.5,0,0.5,1,feather=0.3" --mask-adjust "sky:exposure=-0.7" |
named linear/radial/color/brush masks + AI masks (subject / person / object:car) + combos (combo:A&B); mask_adjust also takes curves/hsl/vignette strings {...} |
| Lens correction | photo-s batch PATHS... --lens-distort 0.15 [--lens-vignette "0.3,0.4" --lens-ca "0.999,1.001"] |
manual distortion / vignette / CA fix |
| Perceptual analysis | photo-s analyze PATHS... --json |
histograms, channel stats, WB lean, exposure, blur - the feedback half of closed-loop grading |
| Rule-based suggestions | photo-s suggest PATHS... [--scale 1.0] [--json] |
analyze stats → conservative fix params (ev/wb_temp/wb_tint/contrast/vibrance/highlight_recovery/levels) each with a reason + the metric behind it; neutral images return an empty dict; zero models, offline - the "decide params" step made automatic |
| AI auto-tone (plugin) | photo-s batch PATHS... --auto-tone 0.8 |
personal-style 9-param prediction (pip install 'photo-s-plugin-auto-tone[model]'); its MCP tools/REST routes register automatically once installed; zero-model alternative: suggest |
| Visual snapshot | photo-s preview IMG [--max-dim 1024] [--json] |
downscaled JPEG + histogram PNG (base64) - pixels for multimodal agents |
| Regional feedback | photo-s analyze PATHS... --grid 4 --json |
per-cell luma/color + sky/skin ratios + over/underexposed boxes |
| Quality gate | photo-s audit PATHS... [--over-max 2 --blur-min 0.05] [--json] |
pass/fail + reasons - the agent's stop condition |
| Version compare | photo-s diff A.jpg B.jpg --json |
PSNR / SSIM / MAD between before/after |
| LR training package | photo-s lr-scan [--export-dir DIR --render-dir DIR --sanitize] |
discover .lrcat/.xmp → coverage report + JSONL labels + before images |
| Auto-tone train | photo-s lr-train --data lr_records.jsonl --images before/ --out m.npz |
ridge regression, pure numpy, 9 global params (needs >=30 edited photos) |
| Auto-tone predict | photo-s lr-predict IMG --model m.npz |
image → 9 global params (auto-detect ridge / CLIP+MLP npz) |
| Recipe clustering | photo-s lr-recipes --data lr_records.jsonl |
your personal style signatures as PhotoS options |
| Similar-edit search | photo-s lr-similar IMG --data lr_records.jsonl --images before/ |
content-feature kNN → past edit as starting point |
| Eval set prep | photo-s lr-eval --data ... --out eval.json --sample 200 |
before/after render pairs + teacher scoring template |
| Trace log | photo-s batch PATHS... --trace DIR |
before-analyze → params → after-analyze per file (training format) |
| XMP export (Lightroom roundtrip) | photo-s xmp-export IMG... [--embed] [--auto-tone 0.8] [--rating 4 --keywords a,b] [--json] |
writes adjustments as LR-readable XMP: JPEG gets it embedded (--embed — LR ignores sidecars on JPEG), RAW/other formats get a .xmp sidecar; masks/rating/keywords included; --auto-tone records the actually-applied predictions; batch variant: batch ... --write-xmp |
| Unattended pipeline | photo-s autopilot DIR [--mode suggest|auto_tone|both] [--scan-existing] [--write-xmp] |
watch → suggest/auto-tone → process → audit gate → route to passed/ review/ + JSONL trace; --write-xmp hands off to human LR fixes (residual training signal); stop with Ctrl+C; MCP autopilot_start/status/stop for async |
| Semantic index | photo-s index PATHS... [-r] [--tags "landscape,night" --min-score 0.2 --write-xmp] |
builds an embedding index (.photo-s-index.npz, incremental by mtime+size); plugin = SigLIP text+image, bare install = 84-dim histogram (image-only); --tags auto-tags top matches into EXIF/XMP keywords |
| Semantic search | photo-s find "query words" [-k 10] or photo-s find --image REF.jpg |
cosine ranking over the index; prefer English queries with SigLIP (Chinese unreliable); mismatched/missing index → explicit error telling you to rebuild |
| Environment probe | photo-s info --json |
version, optional features, plugins |
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.
- yesterday Changed · +16 lines · +97 tokens per session e79d80f89168
- 12d ago First seen · 154 lines · 346 tokens per session scan A 1455b5bc69b1
photo-s is a skill published in the GitHub repository Dongwu259/photo_s (0 stars, last pushed 2d ago), licensed MIT. It adds 443 tokens to every session and 4,004 once invoked, about $0.0022 per session on Opus 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 2026-08-31.
Other skills, from other repositories
gwm
Manage git worktrees across any repository with the gwm Rust binary (CLI + ratatui TUI). Use when the user asks to create / list / remove / bootstrap / switch / link worktrees, run a command across worktrees (gwm exec) or reclaim build artifacts (gwm clean), materialise a PR into a worktree (gwm review), drive a…
bibverify
Verify, repair, explain, and generate BibTeX references with Bibverify's DOI-first CLI and MCP tools.
ppt-orchestrator
A dispatcher for creating presentations, slide decks, long images, and weekly reports. It chooses a visual style and output format, then sends the work to a more specialised add-on.
zhongguose-html-skill
An HTML-only presentation system using traditional Chinese colours and Chinese-style visual design. It helps organise source material into a report structure and produces a self-contained HTML file with its assets.
dashiai-ppt
A presentation-making add-on that creates an HTML slide deck from preset visual page designs. The result can open offline in a browser and can be exported as a PowerPoint or PDF file.
wechat-delivery
A workflow for preparing WeChat Official Account articles, images, covers, and publishing files in Chinese. WeChat Official Accounts are channels used by organizations to publish articles to followers.