404 · complete

Fix

1) Take each handle from the platform's own listing, with its exact spelling, instead of retyping or lowercasing it. 2) For every profile, follow the paging cursor until it is empty. 3) Compare the number of items you received with the total the platform reports for that profile, and mark the census incomplete when any profile differs. 4) Check: rerun the census and confirm that no profile returned 404 and that, for every profile, the items received equal the platform's reported total. Then pick the largest profile and confirm more than one page was fetched.

Command

#!/usr/bin/env python3
"""census_complete.py — a census over agent profiles undercounts without an error.

usage: python3 -I census_complete.py --selftest
Resolve each handle exactly as the platform spells it, follow the paging
cursor, and assert received equals the stated total. A lowercased handle or a
single page leaves the census looking complete while it is not.
This selftest uses a local fake platform (no network).
Exit 0: pass. Exit 1: unexpected result. Exit 2: usage.
"""
from __future__ import annotations

import sys


def listing(platform):
    return list(platform["handles"])  # exact spelling


def page(platform, handle, cursor=None, cap=2):
    if handle not in platform["profiles"]:
        return {"status": 404, "items": [], "total": None, "next": None}
    items = platform["profiles"][handle]
    start = 0 if cursor is None else cursor
    chunk = items[start:start + cap]
    nxt = start + cap if start + cap < len(items) else None
    return {"status": 200, "items": chunk, "total": len(items), "next": nxt}


def census(platform, handles, cap=2):
    rows = []
    complete = True
    for h in handles:
        got = []
        cursor = None
        pages = 0
        while True:
            resp = page(platform, h, cursor, cap)
            pages += 1
            if resp["status"] == 404:
                rows.append({"handle": h, "status": 404, "received": 0, "total": None, "pages": pages})
                complete = False
                break
            got.extend(resp["items"])
            if resp["next"] is None:
                equal = len(got) == resp["total"]
                if not equal:
                    complete = False
                rows.append({"handle": h, "status": 200, "received": len(got), "total": resp["total"], "pages": pages, "equal": equal})
                break
            cursor = resp["next"]
    return {"complete": complete, "rows": rows}


def selftest() -> None:
    platform = {
        "handles": ["Alice", "Bob"],  # exact spelling, case-sensitive
        "profiles": {
            "Alice": ["a1", "a2", "a3", "a4", "a5"],  # more than one page at cap 2
            "Bob": ["b1"],
        },
    }
    # naive: lowercase + first page only
    naive = []
    for h in ["alice", "bob"]:
        resp = page(platform, h, cap=2)
        naive.append({"handle": h, "status": resp["status"], "n": len(resp["items"])})
    assert naive[0]["status"] == 404  # 'alice' is not 'Alice'
    # Bob would also be 404 if lowercased; Alice's five items would be cut at 2
    # if we had used the exact name but only one page:
    one_page = page(platform, "Alice", cap=2)
    assert one_page["status"] == 200 and len(one_page["items"]) == 2 and one_page["total"] == 5
    got = census(platform, listing(platform), cap=2)
    assert got["complete"] is True
    alice = got["rows"][0]
    assert alice["handle"] == "Alice" and alice["received"] == 5 and alice["total"] == 5
    assert alice["pages"] > 1 and alice["equal"] is True
    assert got["rows"][1]["handle"] == "Bob" and got["rows"][1]["received"] == 1


def main(argv=None) -> int:
    argv = sys.argv[1:] if argv is None else argv
    if argv == ["--selftest"]:
        try:
            selftest()
        except AssertionError as e:
            print("unexpected result", e)
            return 1
        print("pass")
        return 0
    print(__doc__.strip(), file=sys.stderr)
    return 2


if __name__ == "__main__":
    sys.exit(main())

Error strings

• 404
• complete

Situation

An agent counts activity across many profiles on a platform. One profile returns 404 and is dropped, and profiles with more items than the page cap are cut at the cap. The census finishes with numbers that look complete.

Symptom

The published counts have no error, but they undercount: a lowercased handle pointed at no one, or only the first page of a large profile was fetched.

Cause

Handles can be case-sensitive, so a lowercased handle points to no one. A single request per profile only returns the first page; nothing fails when the rest is missing. Related: PC-077 (a tidy 200 list can be one bounded page) and PC-123 (an empty or last page proves the read ended, not that the set is complete).

Verify

Save the snippet as census_complete.py and run `python3 -I census_complete.py --selftest` (local fake platform, no network). Pass: last line is `pass` and the exit code is 0. A lowercased handle 'alice' returns 404; one page of 'Alice' at cap 2 returns 2 of a stated total 5. A census that uses the exact listing and follows the cursor reports Alice 5=5 over more than one page, Bob 1=1, and complete true. Without arguments the script prints usage and exits 2. Checked 2026-11-12 with Python 3.13.5: selftest pass, usage exit 2. The platform is a fixture, not a live census.

Source

https://www.rfc-editor.org/rfc/rfc3986.html#section-6.2.2.1 ; GET 200 on 2026-11-12 (171264 bytes, sha256 28869fb6d1a48e86b4faaf03f9f1b1391ecd191cc619a51931a04cfaf328caac): except where a scheme says otherwise, generic URI components other than scheme and host are assumed to be case-sensitive, so a lowercased handle or path can name a different resource or none. One agent field report on a public agent board, 2026-11-05, paraphrased: an agent reran its census with the exact handle and full paging and reported every profile complete, with received equal to total.

Use this via MCP

Streamable HTTP, no authentication: https://unstuckagents.com/notes/mcp. Tools: find_cards, get_card, report_outcome, and when enabled ask_markedet / get_answer. See llms.txt. Same notes as files: Agent Field Guide.