# Tailrocks Improve Reconcile

> >-

- **Type:** Skill
- **Install:** `agentstack add skill-tailrocks-tailrocks-skills-tailrocks-improve-reconcile`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [tailrocks](https://agentstack.voostack.com/s/tailrocks)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** Apache-2.0
- **Upstream author:** [tailrocks](https://github.com/tailrocks)
- **Source:** https://github.com/tailrocks/tailrocks-skills/tree/main/skills/tailrocks-improve-reconcile
- **Website:** https://skills.tailrocks.com

## Install

```sh
agentstack add skill-tailrocks-tailrocks-skills-tailrocks-improve-reconcile
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Improve Reconcile

Synchronize standalone plan status with repository truth. Apply
[`runtime-trust.md`](references/runtime-trust.md) and
[`reconciliation.md`](references/reconciliation.md). The sole writable output is
`plans/README.md`.
Resolve every relative link in this file against the directory containing this SKILL.md, never the plugin skills root.

Pass only the optional repository path and modifiers; a retained `sweep`
selector is invalid. `--deep` re-verifies every indexed row without sampling. `--batch` makes status
selection deterministic and non-interactive while preserving every evidence,
command, compare-and-swap, and no-change rule. Neither modifier grants plan-body,
source, roadmap, Git, or remote authority. Invoke this owner directly; no
routing skill dispatches it.

## Reconcile

1. Bind root, revision, dirty state, index bytes, and every indexed plan. Refuse
   a roadmap item; `tailrocks-reconcile` owns those.
2. For each row, verify plan existence, stamped SHA, current evidence, changed
   paths, dependencies, and status claim. Default may inspect rows whose state
   or evidence changed; `--deep` re-verifies every row without sampling.
   Run target commands only with explicit authority under the same read-only,
   frozen-input, network-off, secret-scrubbed, external-cache, bounded-process,
   TERM-then-KILL, and post-hash contract; otherwise mark proof unavailable.
3. Mark truth only: `TODO`, `IN_PROGRESS`, `BLOCKED`, `STALE`, or `RETIRED`.
   Missing evidence is not completion. A finding fixed independently becomes
   `RETIRED` with fixing SHA. A defective plan becomes `BLOCKED` and routes to a
   new `tailrocks-improve-plan` invocation; never rewrite it here.
4. Reject symlinked or escaping paths. Atomically compare-and-swap one updated
   `plans/README.md`; rollback only owned bytes. On drift, leave bytes unchanged
   and report the conflict.

## Final gate

Exactly one index update or a no-change receipt. No plan body, source, roadmap,
branch, commit, remote, issue, or comment changed. Every row has current evidence.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [tailrocks](https://github.com/tailrocks)
- **Source:** [tailrocks/tailrocks-skills](https://github.com/tailrocks/tailrocks-skills)
- **License:** Apache-2.0
- **Homepage:** https://skills.tailrocks.com

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-tailrocks-tailrocks-skills-tailrocks-improve-reconcile
- Seller: https://agentstack.voostack.com/s/tailrocks
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.