🧠 Does your memory plugin actually remember?

measures: a solution of yours, bound to this task version · intent=existing_solution

tp clones the solution at its pinned commit and runs it with your provider key against the pinned task version, and progress mirrors privately to your account. The site grades this path: tp resolves the pinned task to an admitted revision, so the judge runs here and the score is not self-reported. Reading this page creates nothing. Needs the pinned tp build (0.0.15.dev23 or newer; PyPI's 0.0.14 has no --server).

pinned to trapstreet/trapstreet-tasks@006c72b1b464 · tasks/session_memory_recall

solution pinned to Ruqii/trapstreet-solutions@9b36ff9bb838 · session_memory_recall/claude-code-noplugins

a line without ?version= resolves to the best recorded run's version — this page names it

install the CLI (once)

uv tool install --force "git+https://github.com/trapstreet/trap@7729f718a9c660cf5a20677de495453d4528dcc0"

what to do

WHAT THIS LINK MEASURES: the recorded solution "claude-code-noplugins" at its pinned commit (intent=existing_solution). Do not answer the cases yourself.

Reproduce the trapstreet solution "claude-code-noplugins" on the task "🧠 Does your memory plugin actually remember?" and report it to https://trapstreet.run.

The solution is pinned to git+https://github.com/Ruqii/trapstreet-solutions@9b36ff9bb838f555822261d10d7000a60b8c357e#subdirectory=session_memory_recall/claude-code-noplugins; the task is pinned to https://github.com/trapstreet/trapstreet-tasks at 006c72b1b464f495f3fc9f6e29180527f6c3738b (tasks/session_memory_recall) — task version tkv_jtxwpfkwr9i02kgw, the one its recorded runs used; the recorded run used tp 0.0.14.

Run exactly:

    tp run git+https://github.com/Ruqii/trapstreet-solutions@9b36ff9bb838f555822261d10d7000a60b8c357e#subdirectory=session_memory_recall/claude-code-noplugins --server https://trapstreet.run --trust-remote --allow-unanchored

If the solution's trap.yaml binds more than one task, add --task <alias> where <alias> is the trap.yaml key whose `source` points at https://github.com/trapstreet/trapstreet-tasks@006c72b1b464f495f3fc9f6e29180527f6c3738b/tasks/session_memory_recall — the alias is the author's key, not this site's task id; the site resolves the board from the pinned task, not from the alias. The site grades this run: tp resolves the pinned task to an admitted revision, so the score comes from the site's judge rather than from your own report. If tp is missing, or `tp run --help` shows no --server flag: uv tool install --force "git+https://github.com/trapstreet/trap@7729f718a9c660cf5a20677de495453d4528dcc0" (PyPI's trap-cli 0.0.14 predates --server, live progress and site grading; this server needs 0.0.15.dev23 or newer and will say so with HTTP 426). If it has never been paired with this server: open https://trapstreet.run/cli/authorize, copy the token, then tp auth login --with-token --server https://trapstreet.run — pairing is once, not per run. --server is not optional: without it tp resolves the DEFAULT server, finds no credential for it, and runs with sync silently off. The run prints its URL before the first case — give me that URL. Do not tp submit; that is a separate step I will ask for.

machine-readable descriptorhttps://trapstreet.run/api/v2/launch/solution/sol_b0ae0kgjlwhwcura?version=tkv_jtxwpfkwr9i02kgwThe same facts as this page, as JSON: the same intent, origin, both pins, task and version id, the recorded tp build. No credential in it, ever.

the line that brought you here

Reproduce the trapstreet solution "claude-code-noplugins" on "🧠 Does your memory plugin actually remember?" (intent=existing_solution): open https://trapstreet.run/launch/solution/sol_b0ae0kgjlwhwcura?version=tkv_jtxwpfkwr9i02kgw and follow it exactly — it is pinned to that solution commit and that task version; use my paired tp with --server. Give me the run URL it produces; do not submit.

Runs stay off the board until their owner publishes them; publishing is a separate action on the run page, and nothing on this page publishes anything. A run opened with an account is private to that account. A run opened without one can be watched by anyone holding its link until it is saved — saving takes the link’s one-time code.

Task page: /tasks/session-memory-recall · first run guide