land-the-plane

Converts session-end claims into a receipted ledger with evidence, registrations, and handoffs.

1|3|Updated Mar 9, 2026
One-click install
npx skills add https://github.com/AIKAGRYA/dharma_swarm --skill land-the-plane-aikagrya
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: land-the-plane
Source: https://github.com/AIKAGRYA/dharma_swarm/tree/main/.warp/skills/land-the-plane
Command: npx skills add https://github.com/AIKAGRYA/dharma_swarm --skill land-the-plane-aikagrya

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Sessions often end with claims of completed work that were never verified — phantoms, narrated-as-done summaries, and stale status prose. This Skill turns session close into an audit: every claim of done/fixed/exists/passes must be bound to a checkable receipt, registered as broken, handed off, or retracted before the final summary is written. ## Core Features & Use Cases - Claim harvesting and receipt binding: Sweeps the session transcript for every completion claim and binds each to verifiable evidence such as commit hashes, pytest output, on-disk file paths, or spine receipt IDs. - Disposition routing: Routes unreceipted claims to exactly one outcome — appended to docs/state/BROKEN_REGISTER.md, handed off with an exact next command, or retracted from the summary. - Ledger receipt output: Writes a structured JSON ledger to ~/.dharma/session_ledgers/ and emits a LANDING LEDGER table in the final message that gates what the prose summary may claim. - Use Case: At the end of a coding session where you fixed tests, edited files, and drafted a summary, run the close-out to discover one claimed fix was never verified — it gets retracted instead of shipped as a phantom in the session summary. ## Quick Start Ask the agent to land the plane and produce a receipted session-close ledger for everything claimed in this session.

Frequently Asked Questions about land-the-plane

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify session claims before ending a coding session?▼

Harvest every claim of done/fixed/exists/passes from the session, then bind each to a checkable receipt such as a commit hash, pytest tail with exit code, or a file path confirmed on disk. Claims without receipts are registered, handed off, or retracted.

What counts as a valid receipt for a completed task?▼

A receipt is evidence a stranger could check right now: a pushed commit hash, a pytest run with counts and exit code from this session, a file path confirmed on disk, a spine EvidenceReceipt id, or a gate run with exit code. Saying "I verified it" is not a receipt.

What happens to claims that cannot be verified at session close?▼

Unreceipted claims go to exactly one of three dispositions: appended as a BR-NNN row in docs/state/BROKEN_REGISTER.md if something is broken, handed off with an exact next command if work remains, or retracted from the summary if the claim was a phantom.

Can I run builds or long test suites at close to earn a receipt?▼

No. Only cheap read-only checks like ls, git log, or targeted pytest runs are allowed at close. If verification is expensive, the claim stays unreceipted and must be handed off rather than laundered through new build work.

What are the limitations of the session-close ledger?▼

It cannot mutate governance or kernel surfaces, rewrite BROKEN_REGISTER history, or commit anything under ~/.dharma/. If the transcript is too degraded after compaction to harvest claims honestly, it ledgers only what the diff and command history prove.