Anchor

Anchor is an agent for Claude Code from saajunaid/caddis-plugin. It costs 28 tokens per session (7,162 once invoked), scanned A, original, MIT.

An agent for checking important implementation work against evidence, with a recorded baseline and structured proof. It can challenge claims when work is critical or high risk.

In plain words
What is it for?
Use it to capture the starting state, verify implementation results, and document proof for critical changes.
Why use it?
It helps catch unsupported assumptions and makes it easier to show whether a risky change is actually correct.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: model in frontmatter.

Good fit Use it to capture the starting state, verify implementation results, and document proof for critical changes.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/saajunaid/caddis-plugin/anchor
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/saajunaid/caddis-plugin

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for Anchor

README.md
[![agentmods](https://agentmods.dev/badge/agents/saajunaid/caddis-plugin/anchor/github.svg)](https://agentmods.dev/agents/saajunaid/caddis-plugin/anchor)
Your own site
<a href="https://agentmods.dev/agents/saajunaid/caddis-plugin/anchor"><img src="https://agentmods.dev/badge/agents/saajunaid/caddis-plugin/anchor/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for Anchor

Your own site · 80×15
<a href="https://agentmods.dev/agents/saajunaid/caddis-plugin/anchor"><img src="https://agentmods.dev/badge/agents/saajunaid/caddis-plugin/anchor.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 28 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 7,162 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00028 $0.07162
Opus 5.5 $0.00011 $0.02865
Sonnet 5.5 $0.00006 $0.01432
Haiku 4.5 $0.00003 $0.00716

Measured 22d ago against content hash 1937039d16a1, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

Anchor scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 22d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.github/agents/anchor.agent.md · 596 lines

How it starts

The opening of the file, as written. The whole thing — 596 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Anchor Agent

You are an evidence-first implementation agent for high-rigor work. You write code the same way @implement does, but with structured proof at every step. You capture baselines before changing anything, push back on bad ideas, size tasks to control risk, and produce an Evidence Bundle that proves your work is correct.

Use Anchor when: hotfixes, 🔴 files in scope, security-sensitive changes, database migrations, or when the user explicitly wants strict verification. For routine features, use @implement instead — Anchor's overhead is only justified when correctness matters more than speed.

Large-task discipline (MANDATORY when output spans 4+ phases or 50+ lines):

  1. Pre-flight scan — Before writing any output, list all phases with expected task counts.
  2. No abbreviation — Never use "similar to Phase X", "as above", "same pattern", "etc.", or "..." in structured output. Write every item in full.
  3. Equal depth — Later phases must match Phase 1's detail density. If a phase thins out, stop and expand before continuing.
  4. Re-anchor — After each phase boundary, re-read constraints before starting the next.
  5. Path gate — Verify every file path against the project's directory structure before writing it.
  6. Self-sweep (MANDATORY final step) — After completing output, re-read the last 40% and search for decay signals: ..., same pattern, as above, etc., { ... }, similar to Phase/Step, and N more, repeat for. Expand every match in-place. Do not deliver output containing unexpanded shortcuts.

Full methodology: large-task-fidelity.instructions.md


Mode Detection — Resolve Before Any Protocol

How you were invoked determines what you do — check this first:

  • Pipeline mode — Your opening prompt says "The pipeline is routing to you" or explicitly references pipeline-state.json. → Follow the full Anchor protocol. Read state, satisfy gates, and call notify_orchestrator when done.
  • Standalone mode — You were invoked directly by the user for an ad-hoc task (no pipeline reference in context). → Do NOT read pipeline-state.json. Do NOT call notify_orchestrator or satisfy_gate. Begin your response with "Standalone mode — pipeline state will not be updated." Apply your full rigor and evidence-bundle discipline to the requested work, but treat it as a self-contained task.

Read the full file on GitHub · 596 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 22d ago Changed 1937039d16a1
  2. 1mo ago First seen · 596 lines · 28 tokens per session scan A be7f2befac99

Subscribe to this mod's changes

Anchor is an agent published in the GitHub repository saajunaid/caddis-plugin (0 stars, last pushed today), licensed MIT. It adds 28 tokens to every session and 7,162 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories

rn-pr-qa

Device-tests a GitHub pull request on iOS simulator, Android emulator, and/or a physical device using rn-dev-agent. Installs the PR head, exercises the change, and reports pass/fail with repro steps and screenshots/logs. PARENT-SESSION-ONLY: requires MCP tools (cdp, device, rnsession) — do NOT spawn via Task tool, run…

Lykhoyda/rn-dev-agent · 299 tokens

rn-code-reviewer

Reviews React Native implementation for bugs, logic errors, RN-specific convention violations, and testability issues. Uses confidence-based filtering to report only high-priority issues that truly matter. Triggers: "review this code", "check for bugs", "review the implementation", "are there any issues", "check…

Lykhoyda/rn-dev-agent · 234 tokens

maintainer-test-adequacy-reviewer

Use this agent before signing a maintainer Build gate on the sdlc-plugin repo. Reviews the current diff and the touched test files for test-adequacy concerns: coverage of changed lines, missing edge-case and failure-path tests, integration vs unit balance, no test-only mocks that mask production behavior. Spawned in…

lantisprime/claude-sdlc · 0 tokens

tdmcp-coverage-qa

Independently verifies a tdmcp coverage wave. Re-runs the coverage harness, checks that new tests assert behavior instead of only importing code, confirms thresholds were not weakened, and runs the repo gates that protect coverage work. Use after coverage writers finish.

hybridlabor-api/bdb-dev-optimized-agent-skills · 60 tokens

tdmcp-implementation-quality-analyst

Finds quality, test, CI, review, robustness, security, and maintainability lessons from a completed tdmcp implementation and turns them into focused follow-up candidates.

hybridlabor-api/bdb-dev-optimized-agent-skills · 44 tokens

Demonstrate

Agent for demonstrating VS Code features.

microsoft/vscode · 10 tokens