Is iloop safe to install?

SaferSkills independently audited iloop (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.

CategoryWeightCategory scoreContribution

Securityprompt, exec, net, exfil, eval

35%

100

35.0 pts

Supply chainhash, typosquat, maintainer, lockfile

20%

100

20.0 pts

Maintenancestaleness, pinning, CI

15%

100

15.0 pts

TransparencySKILL.md, perms, README

15%

100

15.0 pts

Communityinstalls, verify, response

15%

100

15.0 pts

Findings & checks · 0 flagged

Securityscore 100 · 0 findings

✓—All security checks passedNo findings in this category for the latest scan.pass

Supply chainscore 100 · 0 findings

✓—All supply chain checks passedNo findings in this category for the latest scan.pass

Maintenancescore 100 · 0 findings

✓—All maintenance checks passedNo findings in this category for the latest scan.pass

Transparencyscore 100 · 0 findings

✓—All transparency checks passedNo findings in this category for the latest scan.pass

Communityscore 100 · 0 findings

✓—All community checks passedNo findings in this category for the latest scan.pass

The primary manifest — the file an agent reads to learn what this artifact does.

configs/skills/iloop/SKILL.md · 1 file

configs/skills/iloop/SKILL.md3.7 KB · Markdown

name: iloop

description: "In-session iterative loop — keeps Claude running in the current session until a task is done. Unlike /loop (which spawns background workers), /iloop stays in the current session and re-prompts each iteration via Stop hook. Use for: 'keep fixing until tests pass', 'iterate until this feature works', autonomous debugging. Triggers on: '/iloop', 'in-session loop', 'keep iterating', 'loop until done'."

argument-hint: '"task description" [--max-iterations N] [--completion-promise "TEXT"]'

user_invocable: true

iloop — In-Session Iterative Loop

Runs Claude in a self-contained loop within the current session — no background processes, no worktrees. Each iteration, Claude reads the last output, checks progress, makes changes, and verifies. The loop continues until the task is done or max iterations is reached.

How it works

User: /iloop "Fix all failing tests"
          ↓
Claude starts → works on task → tries to exit
          ↓
Stop hook detects .claude/iloop.local.md
          ↓
Feeds task prompt back → Claude continues
          ↓
Repeat until: <promise>DONE</promise> | <loop-abort> | max iterations

This is different from /loop (which uses background workers + worktrees). Use /iloop when:

You want to stay in the current session and watch progress
The task needs iterative self-correction within one context
You want Claude to autonomously debug/fix until passing

Usage

/iloop "Fix all failing pytest tests"
/iloop "Implement the auth feature" --max-iterations 25
/iloop "Get CI green" --completion-promise "ALL TESTS PASSING" --max-iterations 30

Signals (output anywhere in response to control the loop)

Signal	Effect
`<loop-abort>reason</loop-abort>`	Terminate immediately (e.g. impossible task, needs human decision)
`<loop-pause>what needed</loop-pause>`	Pause — state saved, resume by reopening session
`<promise>TEXT</promise>`	Signal completion (only when `--completion-promise` set and genuinely true)

Startup

When /iloop is invoked:

Step 1: Run the setup script:

bash ~/.claude/scripts/setup-iloop.sh "$ARGUMENTS" --max-iterations 20 --completion-promise "LOOP_DONE"

Adjust --max-iterations and --completion-promise based on the task:

Simple fix: 10-15 iterations
Feature build: 25-30 iterations
Open-ended improvement: omit --completion-promise, use --max-iterations

Step 2: Tell the user:

▸ iloop started — iterating until done (max 20 iterations)
▸ Cancel: rm .claude/iloop.local.md
▸ Monitor: head -8 .claude/iloop.local.md

Step 3: Begin working on the task immediately. Per-iteration protocol:

Read git log and key files to understand what was done last iteration
Run verification (tests / build / lint) to check current state
Make focused progress — fix one thing at a time, not multiple things
Run verification again after changes — evidence required before marking done
Scan for related issues (fix A → check B and C)
Only output <promise>LOOP_DONE</promise> when ALL criteria are genuinely met

Cancelling

rm .claude/iloop.local.md    # Immediate cancel

Or output <loop-abort>reason</loop-abort> from within the session.

Difference from /loop

	`/loop`	`/iloop`
Execution	Background process, parallel workers	Current session, sequential
Worktrees	Yes, isolated per worker	No, works on main tree
Visibility	Progress via `--status`	Real-time in session
Use case	Large parallel improvements	Focused iterative fixing
Context	Each worker starts fresh	Same context across iterations

iloop— agent skill

Is iloop safe to install?

iloop — In-Session Iterative Loop

How it works

Usage

Signals (output anywhere in response to control the loop)

Startup

Cancelling

Difference from /loop

Audit the pieces. Scan the whole. Decide.