Document ITERATION_CAP and ITERATION_CAP_HUMAN configuration variables in fix agent docs

[ralphbean]

What's missing

The fix agent's iteration cap feature is documented in docs/fix.md under "Iteration limits" (lines 148-159), but the environment variables that control it are not documented in the Variables section. Users can see that caps exist (5 for bot, 10 for human) but don't know how to configure them.

What should be documented

Add ITERATION_CAP and ITERATION_CAP_HUMAN to the Variables table in docs/fix.md, following the same format used in docs/triage.md (lines 148-159) and docs/review.md (lines 86-99).

Variables to add:

Variable	Description	Default	Valid values
ITERATION_CAP	Maximum number of bot-triggered fix iterations per PR. When this limit is reached, the fix agent stops running automatically and applies the needs-human label. Human-triggered /fs-fix commands can still run (up to ITERATION_CAP_HUMAN).	5	Positive integer
ITERATION_CAP_HUMAN	Maximum total fix iterations per PR, including both bot-triggered and human-triggered (/fs-fix) runs. When this limit is reached, all fix runs are blocked.	10	Positive integer

Location: docs/fix.md, Variables section (currently lines 58-62)

Additional work

Add a reference in docs/review.md to point readers to the fix agent's iteration cap configuration, since the review→fix cycle involves both agents. Suggested location: after the Variables section or in the "Control labels" section where it mentions triggering the fix agent (line 52).

Context

The iteration cap prevents infinite review→fix loops when the two agents disagree. The implementation is in scripts/pre-fix.src.sh (lines 97-120), which reads these variables with defaults but they're not surfaced in user-facing docs.

[ralphbean]

Related: #1021 — these variables should be renamed to FIX_ITERATION_CAP and FIX_ITERATION_CAP_HUMAN per ADR 0080 before documenting them.

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 5:48 PM UTC · Completed 5:52 PM UTC

Commit: f20826e · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $0.84

[fullsend-ai-triage]

This issue is blocked on #1021 — as you noted in your comment, the variables should be renamed to FIX_ITERATION_CAP and FIX_ITERATION_CAP_HUMAN per ADR 0080 before being documented. Documenting the current names would create docs that need immediate revision once the rename lands.

Once #1021 is resolved, this issue is ready to proceed — the scope is clear: add the renamed variables to the Variables table in docs/fix.md and add a cross-reference in docs/review.md.

Related: Draft PR fullsend-ai/fullsend#6596 introduces a separate review-level cycle cap (REVIEW_MAX_FIX_CYCLES). If that lands, the iteration-limits section in docs/fix.md may need to mention both cap mechanisms.

Blocked by:

• Rename ITERATION_CAP and ITERATION_CAP_HUMAN to FIX_ITERATION_CAP and FIX_ITERATION_CAP_HUMAN #1021

---

Labels: Documentation issue about the fix agent iteration cap variables, blocked on a prerequisite rename.