mvanhorn/last30days-skill · error · RuntimeError

save_output: could not find a unique filename after 101 atte

Error message

save_output: could not find a unique filename after 101 attempts in {path}

What it means

save_output() tries up to 101 candidate filenames (stem, stem-date, stem-date-1..99) using exclusive create; if every candidate already exists on disk it raises RuntimeError instead of overwriting a prior report. In practice this means the same topic was saved 101+ times into one directory.

Source

Thrown at skills/last30days/scripts/last30days.py:304

                save_root = candidate.parent.resolve()
                if save_root == Path(library.DEFAULT_MEMORY_DIR).expanduser().resolve():
                    library_index.sync_library(save_root)
                else:
                    # A scoped save must sync a per-directory index with the
                    # same paths scoped search uses; syncing the shared DB
                    # from one scope's scan would prune other scopes' rows.
                    library_index.sync_library(
                        save_root,
                        save_root / "briefings",
                        db_path=save_root / ".last30days-library.db",
                    )
            except (library_index.LibrarySearchUnavailable, OSError, sqlite3.DatabaseError):
                # Saving research must not depend on the optional local index;
                # `library search` reports a clear capability error on demand.
                pass
        return candidate
    # Fallback: all 101 candidates existed (extremely unlikely).
    raise RuntimeError(
        f"save_output: could not find a unique filename after 101 attempts in {path}"
    )


def save_rendered_output(
    rendered_content: str,
    output_file: str,
    *,
    private: bool = False,
) -> Path:
    out_path = Path(output_file).expanduser().resolve()
    _ensure_output_directory(out_path.parent, private=private)
    if private and out_path.exists():
        out_path.chmod(0o600)
    fd = os.open(
        out_path,
        os.O_CREAT | os.O_TRUNC | os.O_WRONLY,
        0o600 if private else 0o644,

View on GitHub (pinned to c7460f6114)

Solutions

1. Clean or archive old outputs for that topic in the save directory (or point --save-dir at a fresh directory).
2. Lower run frequency or make the save path unique per run (include hour/minute in the topic slug or save dir).
3. If overwrite semantics are acceptable, remove the stale dated files before re-running.

Example fix

# before
cron: every 15 minutes, same save dir

# after
SAVE_DIR="runs/$(date +%Y%m%d-%H%M)" python3 last30days.py topic --save-dir "$SAVE_DIR"

stem_candidates = 1 + 1 + 99  # engine probes at most 101 names per stem/date
existing = len(list(save_dir.glob(f'{stem}*')))
if existing >= 100:
    raise RuntimeError(f'save dir saturated for stem {stem!r}; archive old outputs first')

try:
    path = save_output(content, out_file)
except RuntimeError:
    out_file = str(Path(out_file).with_name(f"{Path(out_file).stem}-{os.getpid()}{Path(out_file).suffix}"))
    path = save_output(content, out_file)

When it happens

Trigger: Repeatedly running the same topic on the same date into the same save dir (cron/monitor loop), so stem.md, stem-YYYY-MM-DD.md, and stem-YYYY-MM-DD-1..99.md all exist and os.O_CREAT|os.O_EXCL fails FileExistsError for all 101 candidates.

Common situations: A scheduled trend-monitor running many times per day; a retry loop re-saving after partial failures; two concurrent runs racing for the same filenames.

Related errors

• Could not find a unique discovery output filename
• [last30days] Cannot read --synthesis-file: {exc}\n
• [CompetitorsPlan] Cannot read plan file: {exc}\n
• [Planner] Cannot read --plan file: {exc}\n
• No handoff location available to write the nominations bundl

AI-assisted analysis of mvanhorn/last30days-skill@c7460f6114 (2026-08-15). Data as JSON: /api/errors/0835b412e3ecbb81. Report an issue: GitHub.