# Align Project Decisions With Your Expectations

Unclear goals and changing priorities can make agent decisions diverge from your expectations. Align captures confirmed criteria in ALIGN.md and reviews meaningful changes against them.

## Install

```bash
npx skillstore add hmmhmmhm/align
```

## Metadata

- Status: approved
- Slug: hmmhmmhm-align
- Skillstore revision: r1
- Version status: missing
- Tree hash: 57c1770518bfbbb394ed0c794bd5f97b8c275e15c3f890cda5f2444deb54db47
- Author: hmmhmmhm
- GitHub username: hmmhmmhm
- License: MIT
- Repository: https://github.com/hmmhmmhm/align.md/tree/b68ffcfe0e495f43ccd7c33a721aaaf21aa62e42/skills/align
- Ref: 8a2eeb88c8f9499e24713e1fb2caafe4a7568aa5
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/hmmhmmhm-align
- Manifest: https://skillstore.pages.dev/api/skills/hmmhmmhm-align/manifest

## Capabilities

- Review existing agreements, relevant project documentation, conversation context, and available Git history.
- Ask one to three focused questions about consequential gaps in goals, priorities, constraints, and autonomy.
- Draft a concise agreement that separates requirements, preferences, assumptions, and unresolved matters.
- Save authorized agreement changes in the project-root ALIGN.md while preserving unrelated content.
- Distinguish full confirmation timestamps from partial confirmations and unconfirmed drafts.
- Add an ALIGN.md reference to project agent instructions while preserving existing rules and respecting instruction hierarchy.

## Use Cases

- Define a New Project: Clarify intended users, success criteria, priorities, and exclusions before an agent makes implementation decisions.
- Review Development Drift: Compare relevant implementation changes with an existing agreement and propose revisions only when new evidence matters.
- Clarify Writing Priorities: Record audience, quality expectations, constraints, and autonomy boundaries for an ongoing writing project.

## Prompt Templates

### Start an Agreement

```
Use align to clarify goals for [project]. Ask about the most important gaps, then show a concise draft for my review.
```

### Align an Existing Project

```
Use align for [project]. Review relevant documentation and summarize the current direction. Ask what I want to retain or change before drafting ALIGN.md.
```

### Review Possible Drift

```
Use align to compare [recent changes] with our existing ALIGN.md. Use available history, disclose gaps, and propose revisions only for meaningful inconsistencies.
```

### Update a Specific Criterion

```
Use align to update [criterion] to [new requirement]; I authorize this lasting change. Preserve unrelated criteria and the full-confirmation timestamp. Record this partial confirmation.
```

## Limitations

- Reliable drift review depends on available project context and history; missing evidence must be disclosed.
- The skill does not guarantee every agent reads AGENTS.md or invokes alignment on every task.
- Lasting agreement changes require confirmation or clear authorization; automatic review does not count as confirmation.
- The skill supports one user's alignment with their agents, not a multi-party approval workflow.

## Best Practices

- Identify the target project and provide existing agreements, relevant documentation, and available history.
- State concrete priorities, exclusions, and autonomy boundaries when important values conflict.
- Confirm lasting changes explicitly and distinguish partial updates from full agreement confirmation.

## Anti Patterns

- Treating implementation choices or inferred preferences as confirmed user requirements.
- Repeating a full interview when routine work already fits the current agreement.
- Advancing the full-confirmation timestamp after automatic review or a partial change.

## Security Audit

- Audited at: 2026-10-12T00:43:34.629\+00:00
- Summary: All 14 static findings are false positives: Markdown formatting, a read-only Git check, and duplicate-section guidance. No evidence found of prompt injection, credential access, data exfiltration, or malicious execution. The skill requires authorized agreement changes, preserves existing instructions, and prohibits silent hooks or global configuration changes.

## Stats

- Views: 0
- Downloads: 1
- Favorites: 0
- Popularity score: 0