Weekly doc-drift audit found stale status claims in a locally-maintained
Claude Code skill file (not tracked in any repo, so filing here per the
audit's routing rule).
~/.claude/skills/eigenos-kernel-engineer/SKILL.md:13-36
The overview says:
DONE through M12 (2026-07-03): ... M12 devices in EigenScript (...
MMIO deliberately deferred to the first real MMIO device, the M13
NIC). OPEN: M7 T3200 hardware boot (...) → M13 net (the e1000 is
already visible from script: pci.scan of [] shows it as class 0002).
This understates EigenOS's actual state. Per EigenOS's ROADMAP.md and
README.md Status section (and git log in that repo):
- M13-1 (DONE 2026-07-13): the frame seam — e1000 rings in C, ARP
in EigenScript, RX on the tape.
- M13-2 (DONE 2026-07-13): the protocol tower in script — ARP,
IPv4 (+ICMP echo), UDP, and a DHCP client, all four .eigs modules
layered on the M13-1 frame seam. The machine leases an address via
its own DHCP handshake at boot and pings its gateway under slirp.
So M13 has gone well past "the e1000 is visible from script" — the NIC
driver and a working protocol tower (ARP/IPv4/ICMP/UDP/DHCP) are both
shipped. Only TCP remains open within M13 (explicitly deferred per
README.md's Status section, pending the first consumer that needs a
byte stream).
Separately, two more recent EigenOS commits aren't reflected in the
skill's "DONE through" summary at all: d4befd8 (desktop imports
eigen-edit as the 'edit' window) and 6d557a7 (desktop consumes
eigen-sheet as the win_sheet window), both from 2026-07-20/21.
Fix: update the skill's overview paragraph to reflect M13-1/M13-2 done
and the two new desktop-consumed app windows, and reword "OPEN: ... →
M13 net" so it doesn't read as "not yet started."
~/.claude/skills/eigenos-kernel-engineer/SKILL.md:33-36 — the "M13
net" framing is the concrete stale text; new text should name M13-1/
M13-2 as done and TCP as the sole remaining open item.
Also checked and clean: aot-differential/SKILL.md (deliberately avoids
hardcoding a version — reads ouroboros/.devcontainer/Dockerfile's
EIGS_REF live, per its own note "this skill has drifted twice by naming
one") and write-eigenscript/SKILL.md (all version references are
historical "since vX.Y.Z" behavior-change notes, still accurate, not
"latest version" claims).
Weekly doc-drift audit found stale status claims in a locally-maintained
Claude Code skill file (not tracked in any repo, so filing here per the
audit's routing rule).
~/.claude/skills/eigenos-kernel-engineer/SKILL.md:13-36
The overview says:
This understates EigenOS's actual state. Per EigenOS's ROADMAP.md and
README.md Status section (and
git login that repo):in EigenScript, RX on the tape.
IPv4 (+ICMP echo), UDP, and a DHCP client, all four
.eigsmoduleslayered on the M13-1 frame seam. The machine leases an address via
its own DHCP handshake at boot and pings its gateway under slirp.
So M13 has gone well past "the e1000 is visible from script" — the NIC
driver and a working protocol tower (ARP/IPv4/ICMP/UDP/DHCP) are both
shipped. Only TCP remains open within M13 (explicitly deferred per
README.md's Status section, pending the first consumer that needs a
byte stream).
Separately, two more recent EigenOS commits aren't reflected in the
skill's "DONE through" summary at all:
d4befd8(desktop importseigen-edit as the 'edit' window) and
6d557a7(desktop consumeseigen-sheet as the win_sheet window), both from 2026-07-20/21.
Fix: update the skill's overview paragraph to reflect M13-1/M13-2 done
and the two new desktop-consumed app windows, and reword "OPEN: ... →
M13 net" so it doesn't read as "not yet started."
~/.claude/skills/eigenos-kernel-engineer/SKILL.md:33-36 — the "M13
net" framing is the concrete stale text; new text should name M13-1/
M13-2 as done and TCP as the sole remaining open item.
Also checked and clean: aot-differential/SKILL.md (deliberately avoids
hardcoding a version — reads ouroboros/.devcontainer/Dockerfile's
EIGS_REF live, per its own note "this skill has drifted twice by naming
one") and write-eigenscript/SKILL.md (all version references are
historical "since vX.Y.Z" behavior-change notes, still accurate, not
"latest version" claims).