fix(trace): register checkouts that shell commands work in

[jiparis]

A traced session can work in a second checkout through shell commands only, for example in a fresh clone. Before this change, the trace did not see that checkout. Agent hooks run from the directory where the session started, so the clone got no git hooks, and its commits and pushes carried no session.

This change finds the checkout from the command text. The shell hooks read the command and get these directories from it:

• the target of cd or pushd
• the path of git -C
• the folder that git clone or gh repo clone creates

The trace handles each directory the same as a file tool edit there: it installs the git hooks, records the session, and adds the checkout to the checkouts of the session. A hand-written tokenizer reads the command. This change adds no shell-parsing dependency.

Provider support:

• Claude Code: supported.
• OpenCode: supported. The plugin now sends the command text. A plugin installed by an earlier version must be installed again.
• Cursor: partly supported, through the shell hooks from feat(trace): install the Cursor shell hooks #3624. These hooks are verified for the Cursor IDE only. A relative path in the command resolves against the first workspace root.

Known limit: when one shell call clones a repository and also commits in it, that commit does not carry the session. The clone does not exist before the call, so the git hooks are installed only after the call.

Closes #3533 (PFM-7614)

AI disclosure: this change was made with Claude Code.

[chainloop-platform]

PR validation — ✅ 3 passing

Status	Policy	Material	Messages
✅ Passed	pr-min-approvals	pr-info	-
✅ Passed	pr-description-required	pr-info	-
✅ Passed	pr-user-story-linked	pr-info	-

_{View attestation ↗}

AI Session Checks — 🟢 89% · ✅ 0 failing

Avg score	Sessions	Failing policies	Attribution	Files	Lines	Total Duration
🟢 89%	1	✅ 0	100% AI / 0% Human	13	+957 / -176	3h49m36s

🟢 89% — 100% AI — ✅ All policies passing

Oct 14, 2026 16:36 UTC · 3h49m36s · $19.74 · 520 in / 240.2k out · claude-code 2.1.299 (claude-opus-5-5)

_{View session details ↗}

Change Summary

• Adds shell-command checkout detection so work in other checkouts is tracked.
• Extends OpenCode hook payloads and tests to carry command text for shell events.
• Follows up on review findings with parser fixes for subshells, bare cd, env assignments, and README/lint cleanups.

AI Session Overall Score

🟢 89% — Strong implementation and verification; only the up-front planning lagged task scope.

AI Session Analysis Breakdown

🟢 95% · user-trust-signal

No notes.

🟢 93% · solution-quality

🟢 The AI fixed provider and parser logic instead of weakening checks or bypassing hooks. · High Impact

🟢 92% · alignment

🟢 Scope expansions were checked with the user before the AI moved beyond the original provider. · High Impact

🟢 91% · scope-discipline

No notes.

🟢 90% · verification

🟢 Targeted parser and scenario tests were rerun until the new behavior passed. · High Impact

🟡 72% · context-and-planning

🟠 Implementation, OpenCode, PR work, and review fixes proceeded without a shared plan or TODO. · Medium Severity

💡 Before a multi-step provider change, write a short visible plan with phases and stop points.

---

File Attribution

████████████████████ 100% AI / 0% Human

Status	Attribution	File	Lines
modified	ai	app/cli/pkg/action/trace_command_dirs.go	+551 / -126
modified	ai	app/cli/pkg/action/trace_shell_checkout_test.go	+198 / -5
modified	ai	app/cli/pkg/action/trace_command_dirs_test.go	+117 / -25
modified	ai	app/cli/pkg/action/trace_agent_hook.go	+42 / -1
modified	ai	app/cli/internal/trace/claude/hooks_test.go	+13 / -0
modified	ai	app/cli/internal/trace/opencode/hooks_test.go	+8 / -5
modified	ai	app/cli/internal/trace/README.md	+8 / -4
modified	ai	app/cli/internal/trace/opencode/hooks.go	+7 / -4
modified	ai	app/cli/internal/trace/opencode/testdata/plugin_full.ts	+3 / -2
modified	ai	app/cli/internal/trace/opencode/testdata/plugin_tracerun.ts	+3 / -2
modified	ai	app/cli/pkg/action/trace_agent_hook_test.go	+2 / -2
modified	ai	app/cli/internal/trace/provider.go	+3 / -0
modified	ai	app/cli/internal/trace/claude/hooks.go	+2 / -0

---

Policies (4)

Status	Policy	Material	Messages
✅ Passed	ai-config-ai-agents-allowed	ai-coding-session-2256cd	-
✅ Passed	ai-config-no-dangerous-commands	ai-coding-session-2256cd	-
✅ Passed	ai-config-no-secrets	ai-coding-session-2256cd	-
✅ Passed	ai-config-mcp-servers-allowed	ai-coding-session-2256cd	-

Security Checks — ✅ 5 passing

✅ secret-scan

Status	Policy	Messages
✅ Passed	secrets-detection	-

✅ sast-scan

Status	Policy	Messages
✅ Passed	owasp-top10-2025	-
✅ Passed	sast	-
✅ Passed	cwe-top25	-
✅ Passed	cwe-top26-40-cusp	-

Scans not applied (3)

Scan	Reason
vulnerability-scan	no manifest/lockfile changed
github-actions-scan	no workflow files changed
iac-scan	no IaC files changed

_{View attestation ↗}

Security context

This change touches code with 2 recorded security-fix advisories. These are pointers to what past fixes established, not findings in this diff, and they never fail the check.

_{View in Chainloop ↗ · How this works ↗}

---

Powered by Chainloop and Chainloop Trace

[cubic-dev-ai]

All reported issues were addressed across 13 files

<sub>Reply with feedback, questions, or to request a fix.

View guided diff | Re-trigger cubic</sub>