- Home
- Skills
- Agents & Orchestration
- archlinux upgrade check skill
archlinux upgrade check skill
A skill that checks Arch Linux News and the BBS "Pacman & Package Upgrade Issues" forum for anything that might require manual intervention before you run `pacman -Syu`.
Free
Works with the AI Agents you already use
archlinux upgrade check skill
Example session with this skill installed
I'm about to run pacman -Syu. Can you check for any manual interventions or known issues with my pending updates?
- Read your context and instructions
- Compiled the archlinux upgrade check
Upgrade Check Results
Last full upgrade: 2026-08-17 (about 2 weeks ago) - 6 packages to update
Issue Found: zed package - zsh completions broken in 1.16.1-3
Topic: https://bbs.archlinux.org/viewtopic.php?id=314715 (Aug 24, 2026)
The zsh completion file installed by zed 1.16.1-3 defines the completion
for a command named cli, but the binary ships as zeditor, so tab
completion breaks after the update. The report includes the one-line fix
(compdef _cli zeditor in /usr/share/zsh/site-functions/_zeditor).
Action: If you use zsh, expect zeditor completion to stop working after
updating to 1.16.1-3 until this is fixed upstream. Cosmetic; not blocking.
Also worth a look: KDE Frameworks 6.29 breaks Dolphin via baloo
Topic: https://bbs.archlinux.org/viewtopic.php?id=314637 (Aug 18, 2026)
baloo 6.29.0-1 is missing the liblmdb.so.1 dependency, so Dolphin
fails to start with "error while loading shared libraries". The thread
notes that downgrading baloo to 6.28.0-1 works around it.
No other issues found
No manual-intervention news since your last upgrade. Safe to proceed with
pacman -Syu, but be aware of the two items above.
Connects securely to your tools. The creator never sees your data.
About this skill
archlinux-upgrade-check-skill
A skill that checks Arch Linux News and the BBS "Pacman & Package
Upgrade Issues" forum for anything that might require manual
intervention before you run pacman -Syu. Tested on
pi.
Before a system upgrade, Arch users may want to read the latest news and forum posts. Doing that by hand is tedious -- this skill automates it: it figures out what you are about to upgrade, scans the official news and the upgrade-issues forum since your last upgrade, cross-references every post against your package list, and reports only the items that actually affect you.
What it does
- Finds your upgrade window -- parses
/var/log/pacman.logfor the date of your most recentpacman -Syu, and runscheckupdatesto list pending upgrades. - Scans the sources -- scrapes Arch Linux News (14 pages, back to 2002) and the BBS upgrade-issues forum (since inception). HTML is used instead of RSS because the RSS feeds only cover ~9 months (news) and ~25 days (BBS).
- Cross-references -- matches package names from your update list
against each news title and forum topic (title, first post, and recent
replies). Matching is two-tier: full package name first, then a
base-name fallback for hyphenated packages (e.g.
plasmaforplasma-desktop), with a blacklist and length gate to kill noise likelinux/python/archlinux. - Verifies and reports -- the LLM examines the per-package evidence
(
title,first_post,recent_posts, with snippets andmatch_type) to drop false positives (URL matches, generic mentions) and presents the verified findings to the user, with links.
If your last upgrade was more than a year ago, the scan window is capped
to 12 months and the report sets lookback_capped: true, warning you to
do step-wise upgrades via https://archive.archlinux.org/ instead of a
direct pacman -Syu.
Limitations (what it does NOT do)
This skill is a pre-flight check, not a guarantee. It scans two sources (official News + the BBS upgrade-issues forum) and lets an LLM judge the hits. It does NOT cover:
- Other sources -- AUR comments, other BBS boards, IRC, the arch-dev-public mailing list, bug trackers, and non-English community forums. Scanning is limited to archlinux.org News and the BBS "Pacman & Package Upgrade Issues" board (forum id 44); issues discussed only elsewhere are invisible to this skill. (Posts in any language that DO appear on the scanned board are still matched -- package names are ASCII and the verifying LLM is multilingual.)
- Brand-new posts -- the BBS scrape has a small lag; a thread posted hours before your upgrade may not be scanned yet.
- Package names not mentioned at all -- matching is by explicit package-name string in the title, first post, or replies. A thread whose title lacks the name is still caught (the body is scanned too); only a thread that never names the package (e.g. a pure symptom report like "system won't boot" with no "glibc" anywhere) cannot be linked to your update list.
- Partial-upgrade risk -- the skill does not detect that you ran
pacman -Sywithout-u(a common Arch footgun); it only checks news/forum content, not your pacman history. - LLM judgement -- the final RELEVANT / NOT_RELEVANT verdict is an LLM call; it can over-report (flag a routine upgrade as intervention) or under-report (miss an incidental mention). Always read the linked thread yourself before acting.
Before upgrading, still: read the pacman output during -Syu, keep a
backup / timeshift, and know how to rollback (downgrade via
/var/cache/pacman/pkg).
Example run
The interaction below was captured by spawning pi with the skill and a
prompt, using mock data (evals/mock/e1) so it is reproducible offline:
cd archlinux-upgrade-check-skill
python3 scripts/skill_eval.py --model <provider/model> --evals 1 --mock --repeat 1
--mock runs the agent inside a bwrap sandbox that presents a whole fake
machine: a shimmed checkupdates and pacman bound over /usr/bin, the
fixture at /var/log/pacman.log, the saved pages served over https_proxy,
and CLOCK_REALTIME shifted to the fixture's "today". Nothing inside the
sandbox names the harness -- no MOCK_* variable, no fixture path on disk.
See docs/development/system-mock-design.md.
The script writes a sharded JSON report (a slim report.json summary
plus one match_<k>.json per candidate match with the full post text).
The LLM reads the slim report, then verifies each match and reports back,
for example:
## Upgrade Check Results
Last full upgrade: 2026-08-17 (about 2 weeks ago) - 6 packages to update
### Issue Found: `zed` package - zsh completions broken in 1.16.1-3
Topic: https://bbs.archlinux.org/viewtopic.php?id=314715 (Aug 24, 2026)
The zsh completion file installed by `zed 1.16.1-3` defines the completion
for a command named `cli`, but the binary ships as `zeditor`, so tab
completion breaks after the update. The report includes the one-line fix
(`compdef _cli zeditor` in `/usr/share/zsh/site-functions/_zeditor`).
Action: If you use zsh, expect `zeditor` completion to stop working after
updating to 1.16.1-3 until this is fixed upstream. Cosmetic; not blocking.
### Also worth a look: KDE Frameworks 6.29 breaks Dolphin via baloo
Topic: https://bbs.archlinux.org/viewtopic.php?id=314637 (Aug 18, 2026)
`baloo 6.29.0-1` is missing the `liblmdb.so.1` dependency, so Dolphin
fails to start with "error while loading shared libraries". The thread
notes that downgrading baloo to 6.28.0-1 works around it.
### No other issues found
No manual-intervention news since your last upgrade. Safe to proceed with
`pacman -Syu`, but be aware of the two items above.
More docs
For skill eval and development docs, see https://github.com/zhaozihanzzh/archlinux-upgrade-check-skill.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires Arch Linux and checkupdates utility. Forum scan is limited to one year lookback. Does not execute the actual upgrade, only the safety check.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 20 days ago
- Free with an Agensi account
- Software
Arch Linuxpacman-contrib
PythonBash
Git
Permissions
Allowed Hosts