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.
npx skills add christiandoxa/prodex --skill prodex-session-prompt-writegit clone --depth 1 https://github.com/christiandoxa/prodexWrote 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.
[](https://agentmods.dev/skills/christiandoxa/prodex/prodex-session-prompt-write)<a href="https://agentmods.dev/skills/christiandoxa/prodex/prodex-session-prompt-write"><img src="https://agentmods.dev/badge/skills/christiandoxa/prodex/prodex-session-prompt-write/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.
<a href="https://agentmods.dev/skills/christiandoxa/prodex/prodex-session-prompt-write"><img src="https://agentmods.dev/badge/skills/christiandoxa/prodex/prodex-session-prompt-write.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00031 | $0.00729 |
| Opus 5 | $0.00015 | $0.00365 |
| Sonnet 5 | $0.00006 | $0.00146 |
| Haiku 4.5 | $0.00003 | $0.00073 |
Grade A, and why
prodex-session-prompt-write 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 2d 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.
What it actually says
Existing Prodex session bridge
Use prodex s expose from the same canonical workspace as the one plain
prodex s session. The default MCP tools are:
prodex_session_prompt_write({"message":"..."})for Prompt Write;prodex_session_output_read({"cursor":"...","wait_ms":5000})to read later sanitized visible user, assistant, tool, MCP/agent, and session/turn-status events.
For development requests, resolve the existing session first. Use
prodex_session_prompt_write exactly once when one compatible plain prodex s
exists, then read output with the returned prodex_pid, thread_id, and
next_cursor. The unmodified Codex TUI is not required to render an externally
queued follow-up; use Output Read for the authoritative machine-readable result.
A fresh idle session is valid before its first manual prompt;
the bridge verifies its live app-server thread and delivers the requested message
through Codex. Start one prodex_super_start fallback only after an authoritative
no_session result. Never start the existing-session and fallback paths in
parallel, and never treat addressability, ambiguity, stale identity, queue,
source, or verification errors as no_session.
Prompt Write may return an optional output_cursor anchor when the exact
pre-write rollout source is known; fresh no-source sessions may return null.
It may revalidate and retry once only when the control plane proves the first
attempt was rejected before acceptance or failed during preflight. Inspect
recovery_generation, last_prompt_requeued, and requeue_reason in the
result. Accepted and ambiguous writes are never retried.
write_ambiguous means delivery may have happened after a close, timeout, or
malformed response. Never replay it automatically and make no exactly-once claim.
The two tools share one target identity. Require the same live OS user, canonical cwd, plain-session command, Prodex process birth identity, actual Codex writer ancestry/birth identity, Codex home authority, source identity, and thread UUID. Ambiguity, restart, source rotation, or a changed writer is a fail-closed error; never guess from the newest rollout.
Modern Codex thread authority is the writer's open
thread-writer-locks/<UUID>.lock. Legacy authority is exactly one open
rollout-...-<UUID>.jsonl. If both are present, their UUIDs must agree. A
normal persisted session is required after writing. A fresh Codex 0.153.2
session may be held in the writer's live app-server until its first written
message; the bridge verifies that exact live thread before writing and then
revalidates persistence.
Output reads are bounded and cursor-based. Return only existing sanitized
user-visible assistant/tool/status transcript events. Pages may contain bounded
generic gap markers for malformed, invalid-UTF-8, or oversized records, and
[text_truncated] when text is bounded. Impossible recovery returns
recovery_failed. Never expose hidden reasoning, instructions, credentials,
queue payloads, or raw rollout JSON.
Never read a target PTY or /dev/pts, synthesize keystrokes, write SQLite
queue rows, or start another solver/writer.
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.
- 2d ago Changed · +13 lines f050df8ef4ce
- 3d ago First seen · 45 lines · 31 tokens per session scan A 54932086b1b2
prodex-session-prompt-write is a skill published in the GitHub repository christiandoxa/prodex (47 stars, last pushed today), licensed Apache-2.0. It adds 31 tokens to every session and 729 once invoked, about $0.0002 per session on Opus 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 2026-09-06.
Other skills, from other repositories
plano-agent-skills
Best practices for building agents and agentic applications with Plano, including configuration, routing, orchestration, guardrails, observability, and deployment.
plano-advanced-patterns
Design advanced Plano architectures. Use for multi-listener systems, prompt target schema quality, and layered orchestration patterns.
plano-agent-orchestration
Improve multi-agent orchestration in Plano. Use for agent registration, agent listener wiring, and capability-focused agent descriptions for accurate routing.
plano-config-fundamentals
Validate and fix Plano config fundamentals. Use for config versioning, listener types, provider registration, secrets handling, and startup validation failures.
plano-deployment-security
Apply Plano deployment and production security practices. Use for Docker networking, state storage choices, readiness checks, and environment-based secret handling.
plano-observability-debugging
Improve Plano tracing and debugging workflows. Use for sampling strategy, span attributes, and trace query-based root-cause analysis.