Schedule: the command line epic closes #887

Deliverable. Epic #740 closes: every program in the
suite presents one command-line convention, every invariant §14 names has a test, and the specification says what
the code does. Worked by the session named devlore-cli | develop | pair:claude | Schedule: command line.

The problem

Fifteen of the epic's twenty-three leaves are done, and the work got there opportunistically: features were
picked up as they surfaced rather than in a committed order, and the epic's own plan says status: complete
while carrying an open box. A full audit on 2026-09-11 read every source the epic touches — its issues open and
closed, the specification end to end, every plan naming the convention, the invariant tests, the scenario
assertions, and the help text of all 143 commands across the four programs — and found 43 items, frozen as a
closure ledger. Eight are the
open leaves below; the rest are defects with no issue, placements into other epics, and documents that state
things the code does not do. Nothing has been added to that ledger since it was frozen, and nothing will be:
a new finding appends as row 44 and never renumbers.

The goal

The eighteen lanes below, in order, in four pull requests. The ledger's remaining rows
ride whichever pull request owns them. When the last merges, #740 closes and no re-reading reopens it, because
every source is already read.

Before → after

Lanes

Leaves, in the order committed. One pull request closes several lanes. No lane is split once this schedule is
committed; if delivering a row above needs something not in these lanes, the deliverable has changed and we
reconvene to replan.

PR A closes feature #832 and ledger rows 2, 4, 6, 22,
30, 31, 32, 33. PR B closes feature #833 and rows 12,
13, 14. PR C closes #897, #791 #792 and
#793, both under feature
#463 in another epic and placed here by the owner on
2026-09-15, so the command line work ends with the verbs this specification names. PR D closes feature
#834, rows 36 through 40,
and the epic: rows 41 and 42 correct the epic's plan and body, and #740 closes with the last commit.

Explicitly out — other deliverables, not deferred lanes

• Placements, ledger rows 19 through 27: work whose fix is entirely in cmd/internal/cli but which is filed
  under another epic — devlore-cli#780,
  #763, #673,
  #337, #83,
  #518, #436.
  PR A states where each belongs; moving them is their epics' work.
• Ledger row 3, whether devlore-docs, devlore-index and devlore-inventory join the shared root. A
  ruling that would add lanes rather than close them, and §2 already marks them "not yet".
• Ledger row 16, the ! prefix in an interactive flow. Deferred by design since 2026-09-03.
• Ledger row 17, plan-status drift across docs/plans. Belongs to the process schedule.
• Ledger rows 24 and 25, help text that names retired commands or promises unimplemented work, where the
  command belongs to another epic — fifteen lore commands and three star key leaves. writ repo add
  and remove were on this list until 2026-09-15; they are lanes 8 and 9.
• Rows 44 and 45, found while working Command naming and package layout #835 and not yet placed: --output table orders struct columns
  alphabetically where §8 says declaration order; and every document in the store verifies untrusted because
  nothing seeds allowed_signers.
• Long argument names, ruled 2026-09-12. The specification binds -o for --output by design decision 1 and
  CheckNoOwnOutputFlag enforces it. Retiring the short forms is a change to a shipped ruling and is the owner's.

How the report reads this

star gh issues report --by schedule --schedule "command line" finds this issue by the Schedule: title prefix,
reads the lane table, and renders each lane's live state. Edit this body to move a pointer; the edit history is
the schedule's history.

A schedule issue carries no Epic: label and no **Feature:** marker: it is a view of the work, not work
(issue-standards §Schedule).

Amended 2026-09-19 by the owner: lane 8, #897, inserted and the rest renumbered. Found while writing lane 9's retirement test: writ repo add on the renamed tree exits 0 and prints the group's help, because cobra's legacyArgs does unknown-command checking at the root alone. The rename is invisible without it, so it ships in the same pull request.

Amended 2026-09-19 by the owner, second: lane 11, #793, pulled forward from #894 and the rest renumbered. The layer-journey scenario, skipped on develop for want of repo set, ran on #898's first CI and asserted #793 in the same step as #791, on all six legs; the code follows the scenario, in the same pull request.

Amended 2026-09-20 by the owner: lane 12, #822, pulled forward from #894 and the last two renumbered. Ruled after PR C merged: the epic's remaining lanes are help and reference, and a writ deploy that stops at file.mkdir on every virtual machine blocks the end-to-end sequence every pull request runs.

Amended 2026-09-20 by the owner, second: lane 13, #899, inserted; #759 and #787 renumbered. Found planning lane 12: writ deploy's Examples and the guide name policies the flag rejects, and three comments disagree about the floor. Ruled: its own issue, resolved in lane 12's pull request, so #881 stays on the Planner sentence.

Amended 2026-09-21 by the owner: lane 14, #900, inserted; #759 and #787 renumbered. Found running lane 12's Windows sequence: the deploy stopped in the packages subgraph because the winget driver's post-check matched the id case-sensitively against winget's case-insensitive answer, so #822's occupant was never reached. Ruled: its own issue, resolved in lane 12's pull request.

Amended 2026-09-21 by the owner, second: lane 17, #904, appended. The epic closes on lane 16; lane 17 is a ResourceModel bug taken here because it was found running lane 16's own writ deploy and the owner ruled it addressed while fresh.

Amended 2026-09-21 by the owner, third: lane 18, #906, appended: the catalog and the ledger have no description in the architecture, which is how #904 stayed invisible; a documentation task under the resource model epic, taken while fresh.

Amended 2026-09-22 by the owner, fourth: lane 19, #907, inserted into PR E. Found running lane 17's local deploy: with the ledger ladder in place the deploy still took 182 s, because Mkdir stamps a directory it merely found as its product and the trace digests it as one. Ruled a catalog-contents bug whose fix is local to Mkdir; taken in lane 17's pull request because #904's acceptance depends on it.