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.
git clone --depth 1 https://github.com/pablomarin/claude-codex-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/commands/pablomarin/claude-codex-forge/forge-goal)<a href="https://agentmods.dev/commands/pablomarin/claude-codex-forge/forge-goal"><img src="https://agentmods.dev/badge/commands/pablomarin/claude-codex-forge/forge-goal/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/commands/pablomarin/claude-codex-forge/forge-goal"><img src="https://agentmods.dev/badge/commands/pablomarin/claude-codex-forge/forge-goal.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 | $0.00000 | $0.01007 |
| Opus 5.5 | $0.00000 | $0.00403 |
| Sonnet 5.5 | $0.00000 | $0.00201 |
| Haiku 4.5 | $0.00000 | $0.00101 |
Grade A, and why
forge-goal 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 13d 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 — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Forge Goal Composition Contract
This canonical workflow is installed only as .forge/workflows/goal.md. Forge must never install
.claude/commands/goal.md or .agents/skills/goal/SKILL.md; each host retains its native /goal.
Activation
- Read
.forge/local/state.mdand the active workflow. The developer's native/goalinvocation or explicit native Goal request is the human activation; agent-written state never grants autonomy. - For a new objective, create a nonce, stable objective hash, and activation UUID. Set
turn_countto0,activation_countto1, andturn_ceilingto20. For a same-objective reactivation after exhaustion, retain the nonce and validated count, generate a new activation UUID, incrementactivation_count, and set the ceiling to20 * activation_count. A changed objective always gets a new nonce and starts from zero. - Persist the exact goal fields from
state.template.md, then run.forge/hooks/lib/goal-ledger.sh activate --project <root> --state .forge/local/state.md(or the PowerShell twin). The helper must accept the next sequential activation before work continues. - Compose the host's native Goal over this contract only when authenticated qualification proves
every Must behavior. Otherwise report
BLOCKED; never claim reduced parity.
Loop
On every turn:
- Validate the repository binding, activation sequence, objective nonce, durable count/ceiling,
state, and evidence. At the start of every continued turn, first sync the state's
turn_countto the ledger's already-published count. The Stop hook appends the current turn afterward, so state may lag the ledger by exactly one only between Stop and the next turn; any greater lag or count reduction fails closed. - Resume the exact next unchecked durable step. A same-host resume may restore native context; a cross-host resume starts a fresh native session and never claims session transfer.
- Increment the persistent Forge count through hook-owned records. Native session, token, or turn counters may reset; the Forge ceiling and consumed count never reset after interruption or host switching.
- Record measurable checklist progress and the next step.
FORGE_GOAL_STUCK_WARNINGis advisory: inspect progress, run the cheapest safe falsifying check, or surface a blocker; it does not reset the budget or authorize another broad loop. Do not invoke /council as a routine response. During Goal, Council may resolve a genuine non-destructive decision. It must concern product or engineering judgment and be required to continue the active native Goal. If the check produces a deterministic smallest answer, use that instead. Otherwise apply the chairman's verdict and continue. Council also remains available for an explicit developer request or a concrete high-impact architectural fork that remains unresolved after the check. - If
FORGE_GOAL_BUDGET_EXHAUSTEDappears, preserve the exact checkpoint and stop native autonomy. Resume requires a new native/goalinvocation or explicit native Goal request. Same-objective reactivation adds exactly one 20-turn tranche and continues with the existing count; it never starts at zero. - Invalidate only receipts whose candidate boundary changed and rerun the affected final gates.
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.
- 13d ago Changed · +1 lines 4b58121cc48b
- 16d ago Changed · +7 lines cf1944f8c55a
- 17d ago Changed · +12 lines 20b9c5638112
- 1mo ago First seen · 52 lines · 0 tokens per session scan A 515fcf028f2e
forge-goal is a command published in the GitHub repository pablomarin/claude-codex-forge (5 stars, last pushed 9d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,007 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 2026-09-04.
Other commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.
speckit.bug.fix
Apply the remediation from a bug assessment and record what was changed.