ThumbGate Blocked

Show the enforcement record: how many risky actions were blocked vs warned, which gates fire most, and the full feedback → check → rejection pipeline. Each block is a repeat mistake stopped before it spent tokens or did damage.

This skill wraps existing ThumbGate capability and adds no new logic — it reads the live enforcement counters.

Workflow

1. Headline numbers with the gate_stats MCP tool: blocked count, warned count, and the top

gates by hits. (CLI fallback: npx thumbgate gate-stats.)

1. Full picture with the enforcement_matrix MCP tool: feedback-pipeline stats, active

pre-action checks, and the rejection ledger with revival conditions.

1. Summarize for the user: total blocks, most-triggered gates, and anything in the rejection

ledger close to revival.

1. If counts are all zero, say enforcement is wired but hasn't fired yet and point to the

thumbgate-guard skill to promote a rule.

Field-by-field meanings (blocked vs warned, the rejection ledger, revival conditions) are in references/enforcement-fields.md.

Example

Input: "is ThumbGate actually blocking anything? how many tokens did we save?"

Action: call gate_stats for the counters and enforcement_matrix for the pipeline, then:

14 blocks, 3 warns in the last 30 days. Top gate: no-force-push-main (6 hits). Estimated ~9k tokens saved from re-runs avoided. 1 archived gate is one strike from revival.

Troubleshooting

• All zeros but rules exist: enforcement is wired but no gated action has been attempted yet —

this is normal early on.

• All zeros and rules also empty: nothing has been promoted — use the thumbgate-guard skill.
• `gate_stats` / `enforcement_matrix` unreachable: fall back to npx thumbgate gate-stats, then

run the thumbgate-doctor skill to confirm the MCP server and hooks are live.

Quality checklist (self-verify before delivering)

• [ ] I pulled real counters via gate_stats (or the npx thumbgate gate-stats fallback), not estimates.
• [ ] I included the full pipeline via enforcement_matrix, not just the headline block count.
• [ ] I framed each block as a prevented repeat mistake (tokens/damage avoided), with real numbers.
• [ ] On all-zero, I explained why (wired-but-unfired vs nothing-promoted) instead of implying failure.
• [ ] I added no new logic — only read existing enforcement counters.