h-status - SKILL.md Agent Skill

name: h-status description: | Project state dashboard for haft — read-only consolidation of active problems, pending decisions, refresh-due artifacts, open work commissions, recent notes, and module coverage. Make sure to use this skill whenever the user asks "where are we", "what's pending", "what's stale", "project status", "what needs attention", "show me the state", "what's in flight", "what did we decide on X recently", "haft status" — or whenever a session resumes after a break and situational awareness is needed before deciding what to work on. Cheap, read-only, zero commitments. For verifying a single decision use h-verify. For managing commission lifecycle use h-commission. when_to_use: | Operator wants situational awareness or session-resume context. Cheap and read-only, fire freely. argument-hint: "[optional: context name to filter]" allowed-tools: mcphafthaft_query mcphafthaft_refresh

h-status — Project FPF state dashboard

You are surfacing the current FPF state via mcp__haft__haft_query(action="status"). Read-only — no kernel writes (Step 0's scan is a maintenance write, not a state mutation).

Step 0 — Maintenance check (FPF B.3.4 evidence decay)

Before calling status, scan the most recent kernel response in this session for Refresh reminder: N days since last stale scan. If N > 30, OR if no scan is visible in this session's history:

Call mcp__haft__haft_refresh(action="scan") first — do not defer to the operator. The reminder is a signal for the agent to act, not a prompt for the operator to remember.
Fold any new stale or drifted findings from the scan into the status reply you were already about to produce.
If the scan reveals nothing new — say so briefly and proceed.

Surfacing the reminder is the kernel's job; acting on it is the agent's job. Doing nothing is the failure mode this step exists to fix. See CLAUDE.md Critical Reminders — maintenance discipline.

Step 1 — Call the kernel

mcp__haft__haft_query(
  action="status",
  context="<optional context filter>"
)

Or for richer visualization (terminal-friendly board view):

mcp__haft__haft_query(action="board")

Step 2 — Interpret the response

The status payload contains:

Shipped / Healthy — recent decisions that are still active and within valid_until window
Pending — decisions awaiting implementation, baseline, or measure
Unassessed — decisions without recent measurement
Refresh Due — decisions/evidence past valid_until, drift detected, or R_eff degraded; also epistemic debt budget if exceeded
WorkCommissions Need Attention — blocked, stale, or expired commissions
Backlog — active problems without committed decisions
Addressed — recently closed problems with their decisions
Recent Notes — last 5 micro-decisions
Module Coverage — per-module decision density (governed / blind / stale)

The response also includes a navigation strip with available next actions (e.g., /h-refresh).

Step 3 — Present to operator

Surface the response as-is — the kernel formats it already. Highlight items that need operator attention:

Refresh-due items → recommend /h-verify for the specific decisions
Blocked work commissions → recommend /h-commission with action=show and operator review
Epistemic debt budget exceeded → recommend running /h-verify on the highest-debt decisions
Blind modules (no decisions) → recommend /h-frame if upcoming work targets that module
Stale decisions → recommend /h-refresh action=waive (with new evidence) or action=supersede (with replacement decision)

Step 4 — Optional: cross-link to related decisions when context given

If the operator's context mentions a specific file or module, also call:

mcp__haft__haft_query(action="related", file="<file path>")

To surface decisions whose affected_files include that path.

What NOT to do

Do NOT auto-fix anything from this skill. h-status is read-only.
Do NOT silently filter out refresh-due items — the operator needs to see epistemic debt to triage.
Do NOT use this skill for full-text search across decisions — that's h-search via mcp__haft__haft_query(action="search", query="...").
Do NOT call this skill on every turn. Auto-trigger is fine when the operator explicitly asks; constant polling pollutes context with stale state.

FPF spec references

B.3.4 — Evidence Decay & Epistemic Debt (drives refresh-due classification)
F.10 — Three Ladders: Evidence / Standard / Requirement status
VER-02 — Decay (valid_until expiry semantics)
X-WLNK — Weakest-link aggregation for R_eff degradation surfacing

Look up via mcp__haft__haft_query(action="fpf", query="B.3.4").