BUG: instructions — conversations-send skill emits unsupported --content flag

[andrei-hasna]

Reproduction

The installed conversations-send skill prescribes:

conversations send --channel <channel> --content "<message>" --from <agent>

The installed Conversations CLI rejects that form with:

error: unknown option '--content'

Current conversations send --help defines the message as the required positional <message> argument. Retrying with the positional form succeeded and returned message ID 677940.

Expected

The managed skill must feature-detect and use the current positional-message contract, and its example must be executable against the installed CLI.

Acceptance

• Update the canonical skill source, not rendered copies.
• Add a two-sided contract test: the documented positional form succeeds against a stub/current parser, while the stale --content form is not emitted.
• Run skill validation and the Instructions secret scan.
• Land through one reviewed PR and re-render the managed skill homes.

Deduplication

Public GitHub search found no open Hasna issue or pull request matching conversations-send content.

[andrei-hasna]

Additional contract drift after @hasna/conversations 0.5.29 installed: the managed conversations-send skill says a numeric direct-message reply uses --to-agent, but current conversations reply --help requires --session ; --to-agent now fails as an unknown option. The supported form was proven against blocker message 678061 by reading its session_id and replying with --session agent-chief-operations-quintilian-a78b4865, which created message 678234. Please feature-detect both the positional send contract and the current numeric-DM reply scope; do not freeze another version-specific flag.

[andrei-hasna]

Additional root-cause evidence, 2026-08-07:

The installed parser exposes Usage: conversations send [options] with --channel; it has neither --content nor --space.

A source search of current hasna/instructions also finds two reachable stale emitters in scripts/seed-workflow.ts:

• line 45: conversations send --space {{PROJECT_SPACE}} --content "shipped X"
• line 61: conversations send --space {{PROJECT_SPACE}} --content "shipped: description of changes"

So fixing only one installed skill copy would leave the workflow seeder reproducing the same defect. Please cover every canonical source in this repo that emits conversations send, with the existing two-sided parser contract. Rendered homes remain output, not the edit target.

Todos control: 5195e4f9-4f81-4794-86f8-c135c7413ffb (O2-00083).