Agent update watcher

Skill jajupmochi/agent-harness/skills/agent-update-watcher

Use to keep up with the fast-moving agent ecosystem (new/updated CLIs, plugins, skills) WITHOUT burning tokens on constant checks. Declares the sources to watch in a config, checks them only when a minimum interval has elapsed, and reports ONLY what actually changed vs the recorded version — you then decide whether to adopt.From its SKILL.md

Install

npx -y skills add jajupmochi/agent-harness --skill agent-update-watcher

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
runs commandsInstructs the agent to run 1 command, including `python3 scripts/check_updates.py --config sources.json --state state.json --snapshot latest.json --min-interval-days 7`.

SKILL.md

2.2 KB, 441 tokens by cl100k_base, as published. Nobody here has run it

agent-update-watcher

Overhaul task 6. New agent plugins/skills/CLIs appear constantly; checking them every turn wastes tokens and network. This balances availability / recency against cost with a deterministic checker + a frequency guard, and surfaces only real changes.

Use

Copy scripts/sources.example.json to a local sources.json; list what to watch (name, kind, current_version, url) and set real current versions.
Provide latest versions as a --snapshot {name: latest} JSON (the thin fetch that turns each url into a latest version is caller-supplied — kept out of the deterministic checker so it stays testable/offline):
```
python3 scripts/check_updates.py --config sources.json --state state.json --snapshot latest.json --min-interval-days 7
```
It prints one UPDATE\t<kind>\t<name>\t<cur> -> <latest>\t<url> line per changed source, records the check time in state.json, and skips cheaply if re-run within the interval (override with --force).

Why deterministic + interval-guarded

Checking and diffing versions is pure code (no LLM). The interval guard is the token/network balance: a weekly (default) check is enough for an ecosystem that moves in days, not seconds. The model only decides whether to adopt an update — it never does the polling.

Relation to autoresearch-toolfinder

autoresearch-toolfinder tracks ML-RESEARCH tools (awesome-lists). This watches the AGENT ecosystem itself (the CLIs/plugins/skills you build on). Different domains, same "track + diff + report, don't reload" idea.

Status: v0.1 (config + interval-guarded diff, --snapshot-driven, tested test_check_updates.py 6/6). The real release-feed fetch (url -> latest version) + a scheduled runner are the thin wrappers left to add.

What ships with it: 3 files

7.2 KB alongside SKILL.md, 2 of them executable

scripts/

Gives 0 of the 12 instructions most context ai engineering skills give in 441 tokens

Counted across 1,328 of the 2,349 authors here whose files we hold, read 2026-09-06

Dispatch a fresh subagent for each taskin 76 of 1328, across 59 files
Perform spec compliance review before code quality reviewin 44 of 1328, across 34 files
Dispatch a final code reviewer after all tasksin 38 of 1328, across 26 files
Answer subagent questions before allowing implementationin 36 of 1328, across 26 files
Use the least powerful model capable of the taskin 33 of 1328, across 26 files
Create a TodoWrite list for all tasksin 32 of 1328, across 22 files
Perform a task review after each implementationin 31 of 1328, across 24 files
Extract all tasks and context from the planin 29 of 1328, across 20 files
Provide full task text to subagentsin 28 of 1328, across 20 files
Use git worktrees for isolated workspacesin 25 of 1328, across 20 files
Specify the model explicitly when dispatching a subagentin 23 of 1328, across 18 files
Execute all tasks from the plan without stoppingin 21 of 1328, across 16 files

Said here and by no other author read

Copy the example sources configuration file
List sources to watch in the configuration
Set current versions for all watched sources
Provide latest versions via a snapshot file
Decide whether to adopt reported updates

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.