shepherd-agents / shepherd
Code health
This codebase scores 7.7 out of 10 for code health, which we rate good. It also scores maintainability 7.8 and static performance 9.9 out of 10. The three are scored separately and never blended into one number. Risk is concentrated, as it usually is: 114 of 1,841 files are git hotspots, and they average 6.4 — which is where the fixes pay off most.
Do next
1 thing worth doing this quarter.
Worth planning
Delete 113 unused symbols and files (1,541 lines)
Nothing in the graph reaches them, across 40 files; every reader and agent pays to skip them.
- symbols
- 113
- lines
- 1,541
- safe to delete
- Yes(inferred)
Medium effort
Improve what Repowise can see
Add a test coverage report
Without one, Repowise cannot tell tested code from untested code, so every test-related action says “unknown”.
- line coverage
- Unknown, no report
Small effort
- Lines of code
- 378.9K
Recent activity
All commitsCommits
- 53b3f018fix(workspace-control): make the generated-artifact annotation fence robust to PEP 6493mo ago
- 8aac8c61docs: describe staged changes without foregrounding the maintainer workspace3mo ago
- 5d5b6087style: ruff-format workspace.py (internal release-bookend catch; parity restore)3mo ago
- 20c07908ci: Landlock probe workflow + Beat-0 safety sentinel on the per-PR path3mo ago
- 3d69e7e3fix(batch-f): public-scope ruff pass + regenerate API pages for the new exceptions3mo ago
- 9e852957feat: 0.3.0 — apply settlement, the Beat-0 safety bar, registration ergonomics (Batch F)3mo ago
Decisions
- capabilities only accepts CommandRequest; coordinator rejects non-accepted variants. Defensive fallbproposed
- Use tr for uppercase rather than ${VAR^^} (bash 4+) so the script runs on macOS's default bash 3.2.proposed
- T3-callers: the typed ReduceRequest path passes the caller's payload through unmodified. Pre-T3 the proposed
- T1a: workspace driver only declares CommandRequest acceptance in its CapabilitySet; T1c expands acproposed
- Trust-by-default (Part A): the detector still checks the target ref (DEFAULT_GROUND_REF) even when aproposed
- Trust-by-default (Part A): the detector still flags the broken prior-lineage pin, but it is off the proposed
Everything we found
5,750 open
Every detector's open count by area, each with its worst example. An inventory, not a to-do list: the actions above are the part worth doing.
- worst item is critical severity. Code health2,976Untested hotspot · vcs-core/packages/core/src/vcs_core/_vcscore_runtime.py46 findings · 9.6 deducted
- worst item is medium severity. Refactoring400main+1.5 health, S effort · docs/_system/scripts/check_names.py
- worst item is medium severity. Decisions633shepherd/packages/dialect/src/shepherd_dialect/workspace_control/workspace.py519 awaiting review · 114 hotspots ungoverned
- worst item is medium severity. Ownership1,568shepherd/packages/dialect/src/shepherd_dialect/workspace_control/workspace.py83% single-owner concentration, 12 commits in 90d
- worst item is low severity. Dead code173provider_events_from_observationsUnreachable function (25 lines)
Composition
Open the graphExplore this codebase
- Docs0 pages across 12 modules
- ChatAsk this codebase a question and get an answer with its sources.
- Files1,841 files with per-file docs, health and history
- ArchitectureDependency graph, layers, and 17 entry points
- Code healthPer-file scores, 3,420 open findings, coverage and refactoring targets
- Knowledge graphEntities, communities, and the paths between them
- Change couplingFiles that keep changing together, mined from commit history
- CommitsChange-risk ranked history with agent provenance
- ContributorsBus factor, per-file maintainers, and the human/agent split
- StatsSize class, origin, lifetime churn, rhythm and records
- CostsWhat indexing this snapshot cost, by model and by run.