How do I install the Context Anchor skill?

Run npx skillmds@latest add harshsinghmp/context-anchor in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Context Anchor safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Context Anchor?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Context Anchor free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under MIT.

Who published Context Anchor?

harshsinghmp (@harshsinghmp) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Context Anchor by harshsinghmp — Agent Skill

Name: Context Anchor
Author: harshsinghmp

File contents

⚓ context-anchor — Working Reference Snapshot Generator

Drop a compact working reference snapshot in <project-root>/.agents/anchor.md to prevent cascading context drift across long sessions, breaks, or task switches.

When to Use

Before switching tasks or workstreams mid-session.
Before delegating a subtask to another agent.
At the start of a session when previous context is cold or stale.
Whenever you catch yourself re-reading dozens of messages to recall system state.

Quick Reference

Section	Content Requirement	Length Constraint
What's True Right Now	Active state, key decisions made, and ruled-out paths	2–4 bullet points
The Working Reference	Single sentence that an agent can act upon immediately	Exactly 1 sentence
Next Action	Single atomic next step with file and line reference	Exactly 1 task item

Procedure

Step 1: Scan Active Session

Identify:

What was being built / fixed / investigated.
What decisions were finalized and why.
What was tried and ruled out.
The exact next concrete step.

Step 2: Write Context Anchor

Save to <project-root>/.agents/anchor.md:

# Context Anchor — <ISO timestamp>

## What's True Right Now
- [1-sentence state of active work]
- [Key decision made: what and why]
- [What was tried and ruled out]

## The Working Reference
> [One sentence that a fresh agent can act on immediately without asking questions.]

## Next Action
- [ ] `path/to/file.ts:line` — [Exact atomic action]

Step 3: Echo Anchor

Display the anchor to the user for instant alignment.

Pitfalls

Verbosity & Noise: Every line must be load-bearing. Delete filler.
Vague Next Actions: Never write "continue coding"; specify the exact file, line number, and function.
Missing Decisions Rationale: Always include the "why" for architectural decisions so future agents don't revert them.

Verification

Confirm <project-root>/.agents/anchor.md exists and is formatted cleanly.
Confirm Next Action points to an exact concrete file and line number.

---name: context-anchordescription: Drop a working reference at any point in a session to prevent cascading context drift. Use when switching tasks, resuming after a break, or handing off between agents.license: MIT---# ⚓ context-anchor — Working Reference Snapshot GeneratorDrop a compact working reference snapshot in `<project-root>/.agents/anchor.md` to prevent cascading context drift across long sessions, breaks, or task switches.---## When to Use- Before switching tasks or workstreams mid-session.- Before delegating a subtask to another agent.- At the start of a session when previous context is cold or stale.- Whenever you catch yourself re-reading dozens of messages to recall system state.---## Quick Reference| Section | Content Requirement | Length Constraint ||:---|:---|:---|| **What's True Right Now** | Active state, key decisions made, and ruled-out paths | 2–4 bullet points || **The Working Reference** | Single sentence that an agent can act upon immediately | Exactly 1 sentence || **Next Action** | Single atomic next step with file and line reference | Exactly 1 task item |---## Procedure### Step 1: Scan Active SessionIdentify:1. What was being built / fixed / investigated.2. What decisions were finalized and why.3. What was tried and ruled out.4. The exact next concrete step.### Step 2: Write Context AnchorSave to `<project-root>/.agents/anchor.md`:```markdown# Context Anchor — <ISO timestamp>## What's True Right Now- [1-sentence state of active work]- [Key decision made: what and why]- [What was tried and ruled out]## The Working Reference> [One sentence that a fresh agent can act on immediately without asking questions.]## Next Action- [ ] `path/to/file.ts:line` — [Exact atomic action]```### Step 3: Echo AnchorDisplay the anchor to the user for instant alignment.---## Pitfalls- **Verbosity & Noise**: Every line must be load-bearing. Delete filler.- **Vague Next Actions**: Never write *"continue coding"*; specify the exact file, line number, and function.- **Missing Decisions Rationale**: Always include the *"why"* for architectural decisions so future agents don't revert them.---## Verification- Confirm `<project-root>/.agents/anchor.md` exists and is formatted cleanly.- Confirm Next Action points to an exact concrete file and line number.

Frequently asked questions

Context Anchor