feat(skills): cover Alpine, rename bodhi-* to hpc-*, split per cluster - #15
Merged
Conversation
The same skills now serve both clusters this tool runs on, and the bodhi- prefix stopped being true the moment they did. Each hpc-* skill is a short SKILL.md holding the rules shared by both clusters, which detects the cluster and delegates the rest to an alpine.md or bodhi.md beside it — an agent reads the system it is actually on and is never fed the other one's partitions, paths, and quotas. slurm-batch and slurm-discovery stay single-file with per-cluster numbers labelled inline, since almost all of their content is shared. Alpine's facts were scraped from a live compute node on 2026-08-27 rather than from docs, and the Alpine files record their survey commands and date so the numbers can be re-verified when CURC changes them — quotas, the purge window, and the 2026-08-05 partition/QOS renames are policy, not constants. The clusters differ where it hurts, so the skills say so explicitly: Alpine tiers its storage (2 GB /home that nothing may be written to, small backed-up /projects, huge purged /scratch/alpine where all work runs) where Bodhi has one shared /beevol and stages on node-local /tmp; Alpine requires a QOS on every job and pairs each partition with its own QOS family; Alpine runs Lmod where Bodhi runs Tcl modules; and Alpine couples memory to CPUs (MaxMemPerCPU=3840M). --install-claude now copies each skill's whole directory rather than SKILL.md alone, and removes a stale bodhi-* copy when it installs the hpc-* successor, so pre-rename installs do not carry two skills claiming the same job. The asset probe accepts either name so a pre-rename checkout via SINTERACTIVE_SHARE keeps working. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
--install-claude copies whole skill directories now, but make install, make install-system and make nodes still enumerated skills/*/SKILL.md, so an installed share tree would carry the routers without the alpine.md/bodhi.md halves they delegate to — and the router tells the agent to read a file that is not there. Widen the wildcard to skills/*/*.md in both the SKILLS list and the per-node install loop. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
The skills ship on two clusters now, and an agent on Alpine was being taught Bodhi's filesystem — and vice versa. This PR makes the cluster skills cover both systems without feeding either system's agent the other's irrelevant details.
What
bodhi-compute/bodhi-software/bodhi-storage→hpc-compute/hpc-software/hpc-storage, since the bodhi- prefix stopped being true.hpc-*skill is now a shortSKILL.mdholding the shared rules, which detects the cluster ([ -d /scratch/alpine ]vs[ -d /beevol ]) and delegates to analpine.mdorbodhi.mdbeside it. The agent reads only the system it is on.slurm-batchandslurm-discoverystay single-file with per-cluster numbers labelled inline, since almost all of their content is shared./home(2 GB, write nothing) //projects(250 G, backed up) //scratch/alpine(~10 TB, purged, all work) layout;curc-quota; mandatory QOS and the partition↔QOS pairs (post-2026-08-05 names);MaxMemPerCPU=3840Mmemory–CPU coupling onacpu; Lmod/module spider;acompilefor builds. Each Alpine file records its survey commands and date so the facts can be re-scraped when CURC changes them.--install-claudecopies each skill's whole directory (for the per-cluster files), removes a stalebodhi-*copy when installing itshpc-*successor so pre-rename installs don't carry duplicate skills, and the asset probe accepts a pre-rename checkout viaSINTERACTIVE_SHARE.Checks
bash -n,shellcheck --severity=warning,groff -ww, and the JSON check all pass locally.🤖 Generated with Claude Code