PROJECT-CONTEXT

Project instructions for Callwright, a self-hosted Node.js service that validates, composes, and places outbound calls through Retell, then retrieves and learns from their results.

In plain words
What is it for?
Use them when changing the MCP server, call-dispatch core modules, CLI adapters, Retell webhooks, validation, prompt assets, retries, or read-back behavior.
Why use it?
They preserve safety and consistency rules such as dry-run calling, limited retries, minimal personal data, mandatory AI disclosure, and one generic Retell agent.

Agent

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.

agentmods
npx agentmods add agents/topness-msft/callwright/project-context
Clone the repo
git clone --depth 1 https://github.com/topness-msft/callwright
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,734 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
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 $0.00000 $0.01734
Opus 5 $0.00000 $0.00867
Sonnet 5 $0.00000 $0.00347
Haiku 4.5 $0.00000 $0.00173

Measured yesterday against content hash 8f9e141682ae, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

PROJECT-CONTEXT 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 yesterday.

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/PROJECT-CONTEXT.md · 138 lines

How it starts

The opening of the file, as written. The whole thing — 138 lines — stays where its author put it; the contents beside it link to each section on GitHub.

PROJECT-CONTEXT: Callwright

One-liner

Callwright is a self-hosted, single-user Node.js MCP server that validates, composes, and dispatches outbound calls through Retell, then retrieves and learns from their outcomes.

Architecture principles (immutable unless explicitly revised)

  • Reusable behavior belongs in pure *-core.js modules; server.js and CLI files are thin adapters.
  • One generic Retell agent receives scenario-neutral dynamic variables; do not create per-scenario agents or scenario-specific dialogue code.
  • Spoken phrasing is language data in prompt/phrase assets. AI disclosure is mandatory and non-removable.
  • place_call is dry-run by default. PII is minimized, values are shared only if asked, and read-backs expose keys/types rather than values.
  • Automatic retry is bounded and disconnect-based: one retry by default.

Module map / key files

  • server.js:38-901 — HTTP MCP server, tool registration, auth, Retell webhook, health route, and process entry point. This file is over 500 lines; main flow: helpers at 63-116, MCP tools at 118-742, HTTP/webhook routes at 746-876, start at 894-901.
  • dispatch-core.js:151-438 — validate/profile/compose/read-back/dispatch pipeline shared by MCP and CLI.
  • setup-core.js:1-434 — Retell provisioning, prompt/analysis migration, configuration, and setup status.
  • lang-core.js — runtime language add/verify/update/remove.
  • learn-core.js — profile enrichment and new-scenario candidate staging.
  • retry-core.js — bounded disconnect classification and retry policy.
  • notify-core.js — optional per-call SMS notification.
  • outcome-core.js — canonical custom-status-first outcome interpretation.
  • generic-prompt.md, generic-prompt.ja.md, lang-phrases.json — spoken behavior and language-owned wording.
  • paths.js — mutable-state locations and CALLWRIGHT_DATA_DIR.

Load-bearing / high-blast-radius paths

  • dispatch-core.js, server.js, and place_call.schema.json gate paid calls and personal-data exposure.
  • generic-prompt*.md controls disclosure, transfer behavior, payment limits, and recipient-facing speech.
  • setup-core.js and update-prompt.js mutate hosted Retell agents; always pin agent and LLM versions.
  • retry-core.js and server.js webhook handling can create additional paid calls; preserve caps and idempotence.
  • paths.js and volume-resident JSON files contain mutable user state and must survive deploys.

Read the full file on GitHub · 138 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. yesterday First seen · 138 lines · 0 tokens per session scan A 8f9e141682ae

Subscribe to this mod's changes

PROJECT-CONTEXT is an agent published in the GitHub repository topness-msft/callwright (1 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,734 tokens. 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.