Skip to content

fix(cli): hermes install --help text omits its AGENTS.md write#1849

Open
Mohak-Agrawal wants to merge 1 commit into
Graphify-Labs:v8from
Mohak-Agrawal:fix/1846-hermes-help-text
Open

fix(cli): hermes install --help text omits its AGENTS.md write#1849
Mohak-Agrawal wants to merge 1 commit into
Graphify-Labs:v8from
Mohak-Agrawal:fix/1846-hermes-help-text

Conversation

@Mohak-Agrawal

Copy link
Copy Markdown
Contributor

Summary

Closes #1846.

Bare graphify hermes install (no --project) dispatches to the same _agents_install() helper that aider/codex/opencode/claw/droid/trae/trae-cn use to write a graphify section into AGENTS.md (see install.py's dispatch_install_cli(), the shared elif cmd in ("aider", "codex", "opencode", "claw", "droid", "trae", "trae-cn", "hermes"): branch). So it does write to AGENTS.md — confirmed by reading the dispatch code.

But unlike every sibling platform's --help line, which says exactly what it does ("write graphify section to AGENTS.md (Aider)", etc.), hermes's line only said:

hermes install          write skill to ~/.hermes/skills/graphify/ (Hermes)

No mention of AGENTS.md at all — which is the silent-mutation / trust-boundary concern the issue raises. README.md already documents this correctly (graphify hermes install # AGENTS.md + ~/.hermes/skills/ (Hermes)), so this was --help-text staleness specifically, not a behavior bug.

What changed

Updated the two hermes install/hermes uninstall lines in __main__.py's help output to mention the AGENTS.md section, matching the phrasing pattern used by every sibling AGENTS.md-writing platform and matching what README.md already says.

No runtime behavior change — hermes stays grouped with its 8 sibling platforms that write/remove an AGENTS.md section on install/uninstall, which looks architecturally intentional and symmetric (uninstall correctly removes the section too). Happy to also look at making the behavior itself opt-in/declared at write-time if maintainers would prefer that over just fixing the help text — flagging that as an option per the issue's two suggested resolutions, but went with the smaller, non-behavior-changing fix first.

Test plan

  • python3 -c "import ast; ast.parse(open('graphify/__main__.py').read())" — syntax valid
  • Confirmed via direct code read (not just the issue report) that bare hermes install really does route through _agents_install(), same as its AGENTS.md-writing siblings
  • Confirmed README.md's existing hermes row already matches the corrected help text

🤖 Generated with Claude Code

…ify-Labs#1846)

Bare `graphify hermes install` (no --project) dispatches to the same
_agents_install() helper aider/codex/opencode/claw/droid/trae/trae-cn use
to write a graphify section into AGENTS.md — confirmed by reading
install.py's dispatch_install_cli(). But unlike every sibling platform's
--help line (which says "write graphify section to AGENTS.md (Aider)"
etc.), hermes's line only said "write skill to
~/.hermes/skills/graphify/ (Hermes)", with no mention of AGENTS.md at
all — the exact silent-mutation trust-boundary concern the issue raises.

README.md already documented this correctly (line 681: "AGENTS.md +
~/.hermes/skills/"), so this was a --help-text-only staleness, not a
behavior bug. No runtime change: hermes stays grouped with its 8 sibling
AGENTS.md-writing platforms, which is architecturally intentional and
symmetric (same uninstall path removes the section too).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

bug: graphify hermes install silently modifies AGENTS.md

1 participant