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 afaraha8403/balakit --skill deep-deliberationgit clone --depth 1 https://github.com/afaraha8403/balakitWrote 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/afaraha8403/balakit/deep-deliberation)<a href="https://agentmods.dev/skills/afaraha8403/balakit/deep-deliberation"><img src="https://agentmods.dev/badge/skills/afaraha8403/balakit/deep-deliberation/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/afaraha8403/balakit/deep-deliberation"><img src="https://agentmods.dev/badge/skills/afaraha8403/balakit/deep-deliberation.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00094 | $0.03389 |
| Opus 5.5 | $0.00038 | $0.01356 |
| Sonnet 5.5 | $0.00019 | $0.00678 |
| Haiku 4.5 | $0.00009 | $0.00339 |
Grade A, and why
deep-deliberation 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 1mo 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.
How it starts
The opening of the file, as written. The whole thing — 370 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Deep Deliberation
Leading words: option tree, evidence tournament, reversibility, adversarial review, premortem, uncertainty, checkpoint, ground truth, circuit, autopilot, plain question.
Deep Deliberation frames a consequential decision, compares viable alternatives,
tests them against evidence, and adjudicates unresolved risks. The pipeline
produces a decision record. Implementation starts only after Checkpoint 3
approval, or immediately under --autopilot.
User Input
$ARGUMENTS
Parse a pace flag from arguments before treating the rest as the decision:
--circuit, --autopilot, --auto (alias of --autopilot). Strip the flag
from the decision text. If both circuit and autopilot appear, autopilot wins.
Pace modes (optional)
Default is interactive: stop and wait at every checkpoint.
The user must opt in this invocation — a flag above, or an explicit phrase
(circuit, autopilot, take recommended, circuit the rest,
autopilot the rest). Do not infer from urgency. Mid-run, the same phrases
switch mode from the current checkpoint. stop circuit / stop autopilot
drops back to interactive.
| Mode | Intermediate checkpoints (1–2) | Final checkpoint (3) |
|---|---|---|
interactive |
Stop and wait | Stop and wait. Do not implement until a separate ask |
--circuit |
Auto-accept the recommended choice; log and continue | Stop. Ask whether to go ahead with the final recommendation |
--autopilot |
Auto-accept; log and continue | Auto-accept. Treat the recommendation as the way to proceed |
Skip the wait, not the work. Stages, evidence tournament, and adjudication
still run. Emit a compact PACE_LOG of every auto-accepted choice.
Cannot proceed: if there is no unique recommendation (tie, both options invalidated, return to Stage 1 required, degraded delegates below minimum), stop and ask even on autopilot.
Proceed (autopilot, or circuit after the user says go ahead): accept the recommendation as the decision and start implementing it. That is the way forward. The decision record is still emitted first.
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 1mo ago Changed · +123 lines · +34 tokens per session 590d2f238347
- 1mo ago Changed · +9 lines 7fd207443a13
- 1mo ago First seen · 238 lines · 60 tokens per session scan A abce6f4b01af
deep-deliberation is a skill published in the GitHub repository afaraha8403/balakit (1 stars, last pushed 1mo ago), licensed MIT. It adds 94 tokens to every session and 3,389 once invoked, about $0.0004 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.
Other skills, from other repositories
designing-data-intensive-applications
Apply Martin Kleppmann-inspired data-system rules when designing reliability, scalability, consistency, replication, transactions, streams, or schema evolution.
implementing-domain-driven-design
Apply Vaughn Vernon-inspired implementation DDD rules when building aggregates, domain events, repositories, application services, or context integrations.
patterns-of-enterprise-application-architecture
Apply Martin Fowler-inspired enterprise architecture rules when choosing layers, service layers, transaction scripts, domain models, repositories, mappers, or DTOs.
working-effectively-with-legacy-code
Apply Michael Feathers-inspired legacy-code rules when changing hard-to-test code safely with characterization tests, seams, sprout methods, or wrap methods.
a-philosophy-of-software-design
Apply John Ousterhout-inspired software design rules when reducing complexity, designing module boundaries, or reviewing APIs and abstractions.
clean-architecture
Apply Robert C. Martin-inspired architecture rules when separating business policy from frameworks, databases, UI, or infrastructure details.