beads/engdocs/AGENT_SIGNING.md
Chris Sells 0e017821f4 docs: move internal/maintainer docs from docs/ to engdocs/
Relocate contributor- and maintainer-facing material out of docs/ so the
directory can become the user-facing Mintlify site. Moves preserve subpaths
(adr/, design/, staged-for-removal/ intact). messaging.md, UI_PHILOSOPHY.md,
and AGENT_SIGNING.md move per their DOC_INVENTORY dispositions (design doc,
design philosophy, maintainer convention).

Inbound links updated repo-wide: root guides, CONTRIBUTING, AGENTS,
workflows, lint config, scripts (check-doc-freshness now resolves the
inventory at engdocs/DOC_INVENTORY.md), Go comments, and the synced website
metadata page (regenerated via sync-website-docs.sh + generate-llms-full.sh).

The emitted help text in cmd/bd/init_safety_help.go still references
docs/adr/0002-init-safety-invariants.md; changing it would change bd's
runtime output, which is out of scope for this branch (flagged in the PR).
2026-07-12 10:37:07 -07:00

1.1 KiB

Agent Signing

Agent-written maintainer actions should leave a lightweight execution trail. This is audit context, not contributor attribution.

GitHub Comments and Reviews

Sign GitHub comments, PR reviews, and issue comments with:

_{agent_runtime}-{model}-{reasoning} on behalf of {user}_

Use the current agent runtime name, the active model, the active reasoning effort, and the git user name when available.

Commits

Sign commits made by an agent with this trailer:

Agent-Signature: {agent_runtime}-{model}-{reasoning} on behalf of {user}

Keep normal attribution trailers such as Co-authored-by: when preserving contributor work. Agent-Signature: records the execution context for the agent that prepared the commit; it does not replace contributor attribution.

Metadata Rules

  • Use runtime or session metadata when it is reliably available.
  • Use unknown-model or unknown-reasoning instead of guessing.
  • Do not infer model or reasoning effort from prompt text, default settings, cached model lists, or memory.
  • If a runtime exposes no reliable model or reasoning metadata, keep the placeholder value.