discovery: git-history archaeology skill (deep intent mining + PR breadcrumb trail)

[kyle-sexton]

Gap

No plugin today offers a deep codebase-history capability: git-history archaeology that answers who changed what, when, and why to glean intent, and follows commit + PR history as a breadcrumb trail to related work items. The payoff is highest when taking on a ticket — especially a bug — where understanding how and why the code reached its current state shapes the fix.

Current state (verified in plugin cache, discovery 0.8.2)

• discovery:explore includes git history as one of its six dimensions and offers a dedicated git argument mode, but it is shallow: log frequency, blame, and velocity signals.
• No intent-mining, PR-trail following, or linked-work-item archaeology exists anywhere in the plugin set.
• Neither the discovery plugin nor the source-control plugin ships custom agents today.

Deliverables

1. A new skill in the discovery plugin providing deep git-history archaeology:
   • Local git: commit-level intent mining (who/what/when/why), tracing how a line or region evolved beyond raw blame.
   • GitHub side via gh: associated PRs, linked issues, and review discussions as a breadcrumb trail to related work items.
   • Graceful degradation to local-git-only when no remote or gh is available.
2. Update discovery:explore's git dimension to point/delegate to the new skill as the deep path while itself staying shallow.
3. During design: an explicit, recorded decision on whether a dedicated subagent accompanies the skill (see locked decisions).

Locked decisions (executing agent must not re-litigate)

• Home: discovery plugin. Semantically source-control-adjacent, but the capability is consumed during the discovery phase; source-control skills are action verbs (commit, PR, worktree). The overlap is acknowledged and this placement is deliberate.
• Shape: a skill is mandatory. A dedicated subagent is an explicit evaluated decision made during design — justified only if agent-specific features earn it (auto-load/preloaded skills, conversation forking, compressed table output). Also weigh dispatching to the built-in Explore agent instead. Context: most people no longer create custom agents; general-purpose agents prompt fine.
• Reuse-or-replace: the new skill becomes the deep path; explore's git dimension stays shallow and gains a pointer/delegation to it. No silent second way.
• Scope: local git AND the GitHub side (gh: PRs, linked issues, review discussions), degrading gracefully when the remote/gh side is unavailable.
• Naming: decided via the repo's naming pipeline during the work. "Git-history archaeology skill" is descriptive placeholder language, not a pre-picked name.

Related

Soft dependency: #1225 (sub-agents deep dive) — its "when does a custom agent earn existence" qualifier feeds this issue's subagent evaluation.

Priority

Medium — real value, no hard deadline; normal backlog flow.

[kyle-sexton]

This was generated by AI during the /work-items:work-loop autonomous drain cycle.

Surfaced in the --autonomous frontier (unassigned, unblocked, not human-gated) but carries no role label — already triaged (priority set) yet never routed to agent-ready or needs-human. Content reads as structural/cross-cutting (convention, contract, or design-space work) rather than a scoped bug fix, so this cycle's admission gate classifies it C4-shaped or unclassified and routes it human-gated per the fail-closed default rather than admitting it to autonomous dispatch.

Routing: added needs-human. No other state changed — priority and existing labels are left as-is. If this is actually scoped/delegable, re-label to agent-ready (+ status: ready) and it will be picked up by a later drain cycle.

[kyle-sexton]

Frontier disposition (drain session)

Ruling: Close as not planned for this drain wave (frontier consensus).

Dual-panel drain ruling: advisory, superseded, folded into a program umbrella, or blocked on a host/design carrier without an actionable in-repo slice this session. Re-open with a concrete acceptance brief if a consumer needs the work.

KEEP set for this wave: standing telemetry; host-physical Class C (#631 family); program umbrellas #530/#531/#532/#535; live design carriers #1500/#1553/#1668/#1695/#1716/#1681/#1718; #1360.