aider

A setup guide for using Watch Skill, a video-watching and search tool, with Aider, a command-line coding assistant. Aider runs the tool as a shell command rather than through MCP, a standard way for tools to connect to AI assistants.

In plain words
What is it for?
Use it to investigate screen recordings, locate bugs, ask questions about specific video moments, and search an indexed video.
Why use it?
It shows how to pass video reports, timestamps, and frame paths into an Aider conversation without reprocessing the video for every follow-up question.

Agent

Install

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.

agentmods
npx agentmods add agents/oxbshw/watch-skill/aider
Clone the repo
git clone --depth 1 https://github.com/oxbshw/watch-skill
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 489 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00489
Opus 5 $0.00000 $0.00244
Sonnet 5 $0.00000 $0.00098
Haiku 4.5 $0.00000 $0.00049

Measured 2d ago against content hash c8769510bb08, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

aider 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 2d 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.

docs/agents/aider.md · 64 lines

What it actually says

Watch Skill in Aider

Status: documentation-verified ☑ — via the CLI, not MCP.

Aider has no MCP client. The RFC is open and the exploratory pull requests were closed unmerged, so as of v0.86.x there is nothing to paste an mcpServers block into. Anyone offering you one for Aider is describing a third-party bridge, not Aider.

That is fine here, because the MCP server is a wrapper around a CLI that Aider can already run.

Install

uv tool install "watch-skill[standard]"   # or: pipx install "watch-skill[standard]"
watch-skill doctor

watch-skill needs to be on PATH, since Aider will shell out to it.

Use it

Inside an Aider session, /run executes a command and offers its output to the model as context:

/run watch-skill watch "https://youtu.be/..." --detail balanced

Aider asks whether to add the output to the chat — say yes, and the report, its timestamps, and the frame paths become part of the conversation.

Follow-ups do not re-process anything:

/run watch-skill ask <video_id> "when does the checkout total go wrong?"
/run watch-skill search "pricing decision"

A recording of a bug

The case Aider is good at — you have a screen recording of a failure and want the fix in the same session:

/run watch-skill extract bug-report <video_id>

That prints a fileable report with the exact frame and timestamp, which Aider can then act on against the source.

Notes

  • Add --transcript-only when you want the words and not the frames; it skips the visual pass and is much faster.
  • Output is markdown on stdout and progress on stderr, so /run picks up the report without the noise.
  • If Aider ships an MCP client, this page becomes the fallback rather than the route — the same watch-skill serve every other client here uses will apply.
Changes

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.

  1. 2d ago First seen · 64 lines · 0 tokens per session scan A c8769510bb08

Subscribe to this mod's changes

aider is an agent published in the GitHub repository oxbshw/watch-skill (320 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 489 tokens. 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.