befund-portfolio
Grades each repo in a directory of git repositories Red/Amber/Green/Gray by worst-signal-wins, and requires an explicit portfolio directory when cwd is itself a git repo. Use when the user asks to "sweep multiple repos", "assess our whole portfolio", "grade all our projects", or names a parent directory containing several git repositories.
- License 许可
- MIT
- 最近更新
- 2026-11-02
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- befund-portfolio
- description
- Grades each repo in a directory of git repositories Red/Amber/Green/Gray by worst-signal-wins, and requires an explicit portfolio directory when cwd is itself a git repo. Use when the user asks to "sweep multiple repos", "assess our whole portfolio", "grade all our projects", or names a parent directory containing several git repositories.
03 SKILL.md 正文
befund-portfolio
Sweep a directory of repositories and grade each one's befund health.
Step 1: Scope gate -- refuse to infer cwd's parent as the portfolio
python3 -B "${CLAUDE_PLUGIN_ROOT}/scripts/befund_cli.py" portfolio-scope-gate --cwd <cwd> --explicit-dir <user-named directory, if any>
Rule portfolio-cwd-git-repo-check: if cwd is itself a git repository and the user did not
name an explicit portfolio directory, this refuses. Ask the user to name the portfolio
directory explicitly rather than walking up to cwd's parent -- a git repo's parent is not an
implicit portfolio container.
Step 2: Enumerate repos
List immediate subdirectories of the (now-confirmed) portfolio directory that contain a .git
directory.
Step 3: Grade each repo, worst-signal-wins
For each repo, resolve its configured output_dir first:
python3 -B "${CLAUDE_PLUGIN_ROOT}/scripts/befund_cli.py" get-settings --repo <repo>
This reads and returns the repo's settings (including output_dir) without gating on
whether befund is enabled there -- do not use check-enabled for this, since it hard-fails
on any repo where befund is disabled, which is exactly a repo this sweep must still be able
to grade Gray rather than abort on. Then check whether <repo>/analysis/befund/ (or the
output_dir this call returned) has any artifacts. Then:
python3 -B "${CLAUDE_PLUGIN_ROOT}/scripts/befund_cli.py" grade-repo --has-artifacts --has-high --has-medium-or-gaps
(Pass --has-artifacts only if artifacts exist; pass --has-high only if any summary contains
a High-severity finding; pass --has-medium-or-gaps only if any summary contains a
Medium/Low finding or a Ready-with-gaps verdict.) Rule
portfolio-grade-worst-signal-wins: a repo with no artifacts always grades Gray --
this branch is checked first in the grading function and cannot be overridden by any
finding, so a never-assessed repo can never read as falsely healthy (Green). Never
synthesize a placeholder grade for an unassessed repo beyond Gray.
Step 4: Write the portfolio report
python3 -B "${CLAUDE_PLUGIN_ROOT}/scripts/befund_cli.py" resolve-output-path --repo <portfolio_dir> --filename befund-portfolio.html
Note the output lands in the portfolio directory, not any single repo's output_dir --
this is the one skill whose output is not scoped to a single repo's analysis/befund/. Write
befund-portfolio.html with one row per repo: grade, and (for graded repos) a short
summary of what drove the grade; for Gray repos, "not yet assessed" and nothing more.
Read-only constraint
Never use Write/Edit outside the resolved portfolio report path, and never modify anything inside a swept repo.
04 同仓库更多 Skills
toy-check
A skill that checks things in the repository and reports what it finds.
demo-check
Reports which generated files changed since the last generation run. Use when the user asks what drifted. Read-only; never edits a file.
demo-check
Reports which generated files changed since the last generation run. Use when the user asks what drifted. Read-only; never edits a file.
demo-check
Reports which generated files changed since the last generation run. Use when the user asks what drifted. Read-only; never edits a file.
demo-check
Reports which generated files changed since the last generation run. Use when the user asks what drifted. Read-only; never edits a file.
zirkel-verify-assumptions
Checks exactly ONE named assumption against real evidence in at most 3 steps and reports one outcome: confidence raised (>=90), reading changed, or still unresolved. Use to resolve a single load-bearing uncertainty — typically a blocking entry surfaced by zirkel-clarify-scope: "verify this one assumption", "is it actually true that X", "check this before we rely on it". For multiple uncertainties, invoke once per uncertainty — the step budget is never shared.