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 agentmods add skills/langerrr/zforge/feature-resumenpx skills add Langerrr/zforge --skill feature-resumegit clone --depth 1 https://github.com/Langerrr/zforgeWrote 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/langerrr/zforge/feature-resume)<a href="https://agentmods.dev/skills/langerrr/zforge/feature-resume"><img src="https://agentmods.dev/badge/skills/langerrr/zforge/feature-resume.svg" alt="Measured on agentmods" 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 | $0.00023 | $0.00922 |
| Opus 5 | $0.00012 | $0.00461 |
| Sonnet 5 | $0.00005 | $0.00184 |
| Haiku 4.5 | $0.00002 | $0.00092 |
Grade A, and why
feature-resume 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 6d 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 — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Interactive Feature Implementation
Continue a feature plan with the planner implementing in-session and checking in with the user between phases.
Load $zforge:feature-execution. It owns phase state, recovery, and acceptance. This skill supplies the interactive intent only.
Arguments
$1: Feature name (snake_case, or will be converted)
Pre-flight
Same as $zforge:feature-orchestrate: resolve the feature directory, read the overview, context, plan, Doc Map, decision ledger and session log, append this session's row, and note inherited standing flags.
Then scan the phase files and report the current state to the user before doing anything: which phases are complete, which is next, what is REPORTED, PAUSED or INTERRUPTED, and which standing flags are open.
Order of work
- REPORTED — a previous session took a report and never verified it. Run acceptance before anything else; do not resume that phase's agent and do not trust the report.
- PAUSED — answer the question first. Present it with enough context for the user to decide, or resolve it from the plan and say that you did. A
REASON: USAGE_LIMIT_95pause has no question in it — pick up from the phase's## Resume Point. - INTERRUPTED — resume per the skill's recovery procedure. Re-orient against disk before continuing; the phase may be further along than the checklist claims.
- FAILED — present the error and ask whether to fix, skip, or stop.
- READY — start the next phase in order.
Executing a phase
The planner implements directly rather than spawning. The phase file's contract still governs:
- Read the phase file and everything in its
## Required Contextbefore starting. - Work the checklist, marking items as they complete.
- Record decisions in
## Decisionsas they are made — including the ones settled with the user in conversation, which are otherwise lost when the session ends. - Fill each
## Evidence Requiredrow when its command runs, from the artifact that command wrote — the command's captured output for an E row, the method walked and the referent it was walked against for a J row. Name artifacts per run, so a re-run does not overwrite the file an earlier row quotes. Writing rows at the end of the phase, from what the console said, is what makes a correct implementation fail its own acceptance. - Keep
## Files Created/Modifiedcurrent.
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.
- 6d ago First seen · 59 lines · 23 tokens per session scan A e2519eaeb495
feature-resume is a skill published in the GitHub repository Langerrr/zforge (10 stars, last pushed 5d ago), licensed MIT. It adds 23 tokens to every session and 922 once invoked, about $0.0001 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 2026-08-31.
Other skills, from other repositories
loop-setup
Session 0. Sets up the loop in this project - detects the stack, checks which connections actually work (git, database, browser, payments sandbox), asks the few decisions only the owner can make, and writes the .loop/ files. Run this once per project, before any planning or building.
loop-plan
The planning session. Turns a goal into an ordered list of thin complete slices, decides which role and model runs each following session, and writes it all to the plan file. Use at the start of an epic, before any code.
dist-design
This skill should be used when choosing between architectural options for how independently-deployed processes or services will communicate, share state, or scale — stream vs. queue vs. RPC, splitting a monolith into services, a replication/partitioning strategy, or cross-service retry and circuit-breaker design. Runs…
loop
The build loop - locate yourself, investigate the real code, build one slice completely, verify it for real, fix and re-verify, then close and hand off. Use for any session that is implementing something from the plan.
loop-close
Closes a slice or a session - adversarial audit of the diff, ledger entry, selective commit, honest report, and the handoff for the next session. Use before considering anything finished.
loop-verify
Runs this project's verification contract and decides honestly whether something is verified, a gap, or a lie. Use before calling anything done, and whenever a report says "should work".