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 firatcand/forge --skill autopilotgit clone --depth 1 https://github.com/firatcand/forgeWrote 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/firatcand/forge/autopilot)<a href="https://agentmods.dev/skills/firatcand/forge/autopilot"><img src="https://agentmods.dev/badge/skills/firatcand/forge/autopilot/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/firatcand/forge/autopilot"><img src="https://agentmods.dev/badge/skills/firatcand/forge/autopilot.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 | $0.00077 | $0.02671 |
| Opus 5.5 | $0.00031 | $0.01068 |
| Sonnet 5.5 | $0.00015 | $0.00534 |
| Haiku 4.5 | $0.00008 | $0.00267 |
Grade A, and why
autopilot 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 — 212 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/autopilot
/autopilot is the single front door for a NEW forge product: it takes a raw
idea through discovery and the whole spec chain — /forge → /draft-prd →
/draft-spec → /ingest-spec → /decompose → /push-to-tracker — and then
rolls into /deliver, so the product goes from interview to delivering
with the user touching only three things: interview answers, spec approvals,
and parked decisions. Everything between the gates is autonomous; the judgment
AT each gate stays human.
It completes the vision arc: /autopilot is the front half (idea → populated
tracker), /deliver is the back half (tracker band → merged + Done), and
/drive is the per-ticket gauntlet inside /deliver. All three ride native
/goal; none self-loops.
This is a per-turn recipe — NOT a loop (ADR)
/autopilot does all it can in one invocation and then RETURNS. It MUST NOT
self-loop: no sleep, no watch, no while true, no --follow, no
gh pr checks --watch. Whenever the next step is a human gate (a spec to
approve) or an asynchronous wait (CI), it reports the state and returns. The
repeat-until-done engine is native /goal — run /autopilot and, once it hands
off, pair the delivery with /goal "<product> delivered". /autopilot is the
recipe; /goal is the engine. (This is the autopilot ADR — see
spec/decisions / the native-/goal ADR.)
Precondition — a forge-initialized repo
The chain ends at /push-to-tracker, which needs .forge/settings.yaml with a
tracker: block. So /autopilot requires a forge-initialized repo. If
.forge/settings.yaml is absent, STOP and instruct the user to run forge init
first — do NOT auto-run the interactive initializer.
Re-entrancy — resume from APPROVED-stage state, never from bare artifacts
Because /goal (or the user) re-invokes /autopilot across turns, it must
resume from the furthest approved stage. The subtle trap: the chain skills
write their artifact BEFORE their review gate passes (/forge writes
spec/BRIEF.md then prints "Gate 1 — review the brief"; /draft-prd writes
spec/PRD.md then "Gate 2"; /decompose writes plans/phases.yaml then
"Gate 4"). So artifact existence ALONE must never skip a stage — that would
resume PAST an unapproved spec and break the never-auto-approve guarantee.
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 First seen · 212 lines · 77 tokens per session scan A 63ddd4f5a738
autopilot is a skill published in the GitHub repository firatcand/forge (13 stars, last pushed 2mo ago), licensed MIT. It adds 77 tokens to every session and 2,671 once invoked, about $0.0003 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 2026-08-30.
Other skills, from other repositories
github-triage
Triage GitHub issues through a configurable label-based state machine. Use when user wants to triage incoming issues, prepare issues for an autonomous agent, or move an issue between workflow states. Repo inferred from git remote; all GitHub calls go through gh.
to-issues
Decompose a plan, PRD, or spec into independently-grabbable vertical-slice issues (markdown file by default, GitHub issues via flag). Trigger when the user wants implementation tickets, work decomposition, or to convert an implementation plan into parallelizable work. Takes a plan file and emits atomic vertical slices.
loop-build
(loop-engineering) Implements exactly one feature from the BACKLOG.md Backlog — syncs main, acquires the loop lock, creates a feature branch, implements all acceptance criteria with tests, visually inspects UI changes, passes the independent verifier, and opens a PR. Use for Build-mode loop runs, or when asked to…
loop-report
(loop-engineering) Runs the daily Report-mode triage — checks open PRs, CI status and the BACKLOG.md backlog, then updates STATE.md and appends to the run log without writing any code. Use for scheduled daily loop runs, project health checks, or when asked to "triage", "run report mode", or "update the loop state".
to-issues
Decompose a plan, PRD, or spec into independently-grabbable vertical-slice issues (markdown file by default, GitHub issues via flag). Trigger when the user wants implementation tickets, work decomposition, or to convert an implementation plan into parallelizable work. Takes a plan file and emits atomic vertical slices.
github-triage
Triage GitHub issues through a configurable label-based state machine. Use when user wants to triage incoming issues, prepare issues for an autonomous agent, or move an issue between workflow states. Repo inferred from git remote; all GitHub calls go through gh.