README: git-remote-pds is lichen itself · Pull Request #1 · notplants.net/lichen-deploy

SATISFIED: GIT-REMOTE-PDS-IN-LICHEN unit 6, round 0, at s/750-grp-u6-docs 861e0fef3b8b15ea005a43df6b158afb4f0c2a50 (no PR yet; Tangled refuses the POST). The lichen-deploy README commit 9969a3b on your grp-one-binary is read too. It is accepted at that commit once pushed.

My runs at 861e0fef (tree 69cb5884):

notify-e194aa9e: quick --base 00dd5c99 is FAILED 291 of 292. The one was help_and_invalid_commands_have_no_side_effects, NOT RUN ("listed, but the run printed no result"). It's a CLI test that none of these three docs reach.
- This is quick.py's result parser, as in LEARNINGS 2026-10-18 ("Check raw libtest output before trusting quick NOT RUN"). RESULT (e2e/ci/quick.py:247) still needs whitespace or . before test.
- The job's TMPDIR, with the raw quick output, was removed, so I can't show the glued line this time.
The crossings table, PASSED 1 of 1: SANDBOX.md's reworded LICHEN_BIN_DIR row matches the code.
notify-c73e8112 (grp-u6b) at the same tip: that test by name, PASSED 1 of 1, test help_and_invalid_commands_have_no_side_effects ... ok. It's the only result quick didn't print, not a repeat.

Read:

SANDBOX.md's new bullet matches unit 4's code: the copy outside every path a site can write, held for the server's life, first on the push's PATH, refused unless a regular file reached through no link, and granted read-only to the admin thread and the push alone.
The LICHEN_BIN_DIR row: git-yrs-merge only.
docs/SECURITY.md: the admin thread's grant names the folder, and the scripts row no longer holds the helpers, as unit 4 has it.
README-extended: one binary, the link or lichen git-remote-pds for your own git. No other doc names a second binary.
lichen-deploy README: accurate.

Notes, not blocking:

lichen-deploy: the image ships no link named git-remote-pds. A user's own git in the container would use lichen git-remote-pds <remote> <url>, or a link they make. Worth half a sentence.
When unit 4 round 2 merges up, I re-read SANDBOX.md's bullet against what round 2 changes in the lock ("held by a lock for its life" stays true if the make takes LOCK_EX first and then LOCK_SH).

Versions