Execution Log: long agent output reaches Slack as a digest instead of a hard truncation

Append-only log of progress for the user's visibility.

Phase transitions

Phase	Entered	Reviewed/approved by	Notes
phase-selection	2026-09-11	—	Tier 3 (human-approves-pr; below humanSignOffMinTier: 4): a rendering change inside the Slack channel, one additive enum key in both schema copies (an autonomy.sensitivePaths entry, as issue-325's reactions block was), no new call, grant, token, scope or state. Brainstorming skipped: the ticket's five numbered asks are the requirement. No authorized the-loop execute reaches this cloud session, so the selection is recorded here and every phase is walked
requirements-definition	2026-09-11		requirements.md — five requirements, five abuse cases
design	2026-09-11		design.md — channels/digest.py, fit in the renderer, the key; decision-118
test-planning	2026-09-11		testing-plan.md — thirteen rows, seven applicable
tasks-breakdown	2026-09-11		tasks.md — six tasks
implementation	2026-09-11		On claude/github-issue-338-lk8282
verification	2026-09-11		evidence/verification.md — rows T1, T2, T7, T8, T10, T12; evidence/security-review.md — five abuse cases, five closed
needs-review	2026-09-11		PR raised; awaiting the owner (tier 3: human-approves-pr)
complete

Pull requests

PR	Scope / tasks	Status
#345	tasks 1–6: the whole work item	open

Progress entries

2026-09-11 — implemented, verified, ready for review

• Phase: implementation → verification → needs-review
• Did: tasks 1–6, red first (test_channels_digest.py did not import against 3b7563c; the three scenarios failed on the raw section text). channels/digest.py: strip_comments (a forward scan), to_mrkdwn (the line and inline rules, applied outside code fences and spans, broadcasts neutralised everywhere), shorten_paths, condense (the block parser, the ask, the numbered lists, the pointers, a two-pass budget — whole when it fits once drawn, else with the footer reserved — the sentence / clause / space cut), truncate (13.10.0's _cap), fit (the one entry point; the raw length is the threshold, the drawn length is re-checked so the one growing rule never posts over the cap). channels/slack.py: long_messages on the config (unknown → digest with a warning), render_blocks(long_messages=) routing the text and excerpt sections through fit, post building the plain-text fallback from the drawn section when the text was over the cap, _cap an alias of truncate for the verbose context lines. commands/channels_cmd.py: the longMessages: line. Both schema copies (one additive enum key, maxChars restated), the two config templates, the option page, the guide (Reading it on a phone, a Limits bullet), the command page, the capability doc, the README and the collaboration reference; decision-118.
• Checkpoint/tests: make check — see evidence/verification.md. New tests: 41 unit (test_channels_digest.py), 3 scenarios (test_channels_integration.py). Existing assertions changed: two — test_bus.py's cap pin now names long_messages="truncate" (and pins the digest beside it), and test_channels.py's schema-default pin gained the new key.
• Self-review: three passes over the diff. Pass one (the code, adversarially) found that to_mrkdwn drew a # comment inside a shell fence as a heading and a ** inside a code span as bold — now every drawing rule runs outside fences and spans (_outside_code); that a Slack link's site-absolute URL (</docs/a/b/c|t>) was shortened as a path — < joined the lookbehind; that a heading that is the question left an empty ** line behind — an emptied block renders nothing; that the broadcast rewrite could grow a 2 900-character text past Slack's limit — fit re-measures the drawn text; that a Java / Node at …(…) line started a trace on any parenthesised indented line — a file:line location is now required; and that a test helper lived in the module — moved to the test. Pass two verified five design claims against the code: the threshold is the raw length; nothing but slack.py imports the digest (pinned by a test on import lines); the footer is the event's URL only; the pipeline's order and the ledger record are untouched (the scenario asserts no record); the lockfile refresh uv run made was incidental and is not in the change. Pass three read the docs against the code: the guide's and the design's before/after examples were the sketch, not the output — replaced with the real digest of the checklist at maxChars: 700; the design's budget prose said the footer was always reserved — now describes the two passes; the option page's as before 13.11 became 13.10.0's behaviour. Nothing else new.
• Next: the owner's review.
• Blockers: none.

2026-09-11 — spec chain drafted

• Phase: requirements-definition → tasks-breakdown
• Did: read the ticket; read channels/{slack,base,publishers,bus}.py, commands/channels_cmd.py, graph/hooks/{sideeffects,selection}.py, the channel test suites, the guide, the option and command docs, the issue-337 spec and decisions 103, 111, 116, 117 at 3b7563c (13.10.0). Established that every text section is cut by one function (_cap) in one place (render_blocks), that GitHub markdown reaches Slack unconverted (the markers literally — the issue-337 log's observation), and that the post's plain-text fallback is uncapped. Settled the three design questions in decision-118: a structural digest the channel computes rather than a model summary; one enum key (longMessages: digest | truncate) with maxChars as the threshold; mrkdwn rendering on every message, the digest only above the cap. Wrote the four artifacts and the decision.
• Checkpoint/tests: baseline — test_channels.py green at 3b7563c (101 passed).
• Next: task 1 (the digest module), red first.
• Blockers: none.

Verification results

Only when this work item declared test-planning away. It did not: results live in testing-plan.md.

What was verified	Command	Outcome	Evidence
—	—	—	see testing-plan.md

Design critic review

Not selected for this work item.

Round	Critic (<harness>/<model>)	Outcome	Findings → disposition	Link

Review cycles

Cycle	Type (self/critic/security)	Reviewer	Outcome	Findings → disposition	Link
1	self	the-loop (this session)	six new findings	code drawn as prose — every rule now runs outside fences and spans; a Slack link URL shortened as a path — < in the lookbehind; an emptied heading left ** — renders nothing; the broadcast rewrite could grow past the cap — re-measured; an over-eager trace start — file:line required; a test helper in the module — moved	this log
2	self	the-loop (this session)	zero new findings	five design claims verified against the code (the raw threshold, the single importer, the footer's source, the untouched pipeline, the incidental lockfile)	this log
3	self	the-loop (this session)	zero new findings (converged)	the docs read against the code: two examples and two sentences corrected	this log
—	critic	—	unavailable — reviews.critics is empty in this repository's config; does not count toward criticReviewCount	—	—
4	security	the-loop checklist	pass; no human sign-off at tier 3	A1–A5 closed	evidence/security-review.md

Security review (gate)

• Mechanism: the-loop checklist (security.review.mechanism: auto; no security-review skill is invocable from this session's plugin set)
• Outcome: pass — evidence/security-review.md, five abuse cases closed
• Human sign-off: not required (tier 3 < humanSignOffMinTier: 4); the owner's PR approval is the gate

Final validation evidence

Requirement	Proof
R1.1	test_a_long_text_is_digested_within_the_limit (five limits, two fixtures), test_render_blocks_digests_the_text_section, test_render_blocks_digests_the_excerpt_too; Scenario: A long agent comment reaches Slack as a digest that leads with the ask; Scenario: A notification's artifact excerpt is digested too
R1.2	test_the_cut_falls_on_a_sentence_boundary, test_a_sentence_too_long_for_the_room_is_cut_at_a_clause_then_a_space
R1.3	test_the_footer_names_the_link_or_the_ticket, test_the_footer_links_the_event_never_the_text, test_nothing_is_left_out_when_everything_fits_compactly
R1.4	test_truncate_is_the_13_10_0_cut, test_render_blocks_truncate_keeps_the_old_cut, test_an_unknown_long_messages_value_resolves_to_digest, test_bus.py::test_render_blocks_caps_text_and_points_at_the_link
R1.5	test_the_fallback_text_carries_the_digest; the first scenario's fallback assertion
R2.1	test_the_first_question_leads, test_a_reply_instruction_leads_when_nothing_asks, test_without_an_ask_the_first_sentence_leads, test_a_question_in_a_heading_leads_and_leaves_no_empty_heading
R2.2, R2.3	test_a_list_becomes_numbered_lines_with_its_boxes, test_the_authors_order_is_kept_after_the_ask
R3.1	test_fences_tables_and_traces_become_pointers, test_code_is_drawn_as_written
R3.2	test_absolute_paths_are_shortened_and_urls_untouched
R3.3, R3.4	test_html_comments_never_reach_slack, test_to_mrkdwn_draws_markdown_as_slack_does, test_to_mrkdwn_is_idempotent_and_leaves_prose_alone, test_a_short_section_is_drawn_and_nothing_else
R4.1	test_a_short_text_passes_through_untouched; Scenario: A short comment reaches Slack untouched
R4.2	test_the_threshold_is_the_raw_length, test_a_text_that_grows_past_the_cap_when_drawn_is_digested, test_long_messages_is_parsed_and_defaults_to_digest, test_channels.py::test_config_defaults_match_the_schema, test_config_schema_parity.py
R4.3	test_status_prints_the_long_messages_line (both values)
R5.1	test_docs_parity.py (p3–p5 on channels.slack.longMessages); the docs table below
NFR faithful, linear	test_every_digest_line_is_the_authors, test_a_pathological_comment_digests_in_linear_time, test_empty_and_odd_inputs_never_raise
A1–A5	evidence/security-review.md

Capability docs

Capability doc	What changed	History row
channels.md	the rendering bullet (drawn as mrkdwn; above maxChars, per longMessages); a new long text is digested bullet (the digest's rules in EARS, truncate, the whole-text rule under the cap, the fallback text, the mrkdwn drawing and the broadcast rule, the status line); two design links	issue-338 row
standing-sessions.md, control-plane.md	unchanged — a standing session's thread is rendered by the same channel and gains the same digest; the control plane is not touched	—

Documentation

Document	What changed
docs/config/cli/channels-options.md	slack.maxChars restated as the threshold and the phone's lever; a new slack.longMessages section (type, default, what digest does and never does, truncate, the whole-text rule, the mrkdwn drawing); the example YAML
docs/guide/slack.md	a new Reading it on a phone section (the four rules, the checklist digest at maxChars: 700, the notification preview, maxChars as the lever, longMessages: truncate, the markers); a Limits bullet (structural, not a summary)
docs/cli/commands/channels.md	the status bullet names the longMessages line
README.md	the channels poll line says long text arrives as a digest
skills/the-loop/reference/collaboration.md	the channels paragraph names the digest
skills/the-loop/templates/cli-config.yaml, .the-loop/cli-config.yaml	the maxChars comment and a longMessages line
.the-loop/cli-config.schema.json, cli/the_loop/schemas/cli-config.schema.json	longMessages (enum, default digest); maxChars's description (byte-identical copies)
docs/decisions/decision-118.md, decisions.md	the decision and its index row
skills/the-loop/SKILL.md, reference/workflow.md, reference/automation.md	unchanged — the operating model itself did not change (what Slack shows is the channel's rendering; the ledger is still the record)