lore

Name: lore
Author: specstoryai

68installs80forks

How to install lore?

specstoryai/getspecstory/lore

$npx -y skills add specstoryai/getspecstory --skill lore

Installs into the current project.

Summary

SpecStory Lore - mine your SpecStory coding histories (any agent - Claude Code, Codex, Cursor, Gemini, and more) into a persistent corpus, surface your reproducible workflows with corroborated evidence, and interactively forge the chosen ones into skills installed across all your

Files · 1

View on GitHub

SKILL.md

1	# Lore
2
3	Your sessions are your lore. This skill turns a user's real coding history into installed skills.
4	A deterministic engine (`scripts/mine-skills.mjs`) parses their SpecStory transcripts - from **every
5	agent** SpecStory captures (Claude Code, Codex CLI, Cursor CLI, Gemini CLI, Factory Droid, DeepSeek,
6	Antigravity, ...) - into a persistent corpus of beats, and returns corroborated candidates.
7	You - the calling agent, whichever harness you are - supply all judgment: name them, discard the
8	generic ones, and interactively forge the good ones into `SKILL.md` packages grounded in the user's
9	own commands.
10
11	The engine does the retrieval and counting; you do the synthesis. Do not try to read raw
12	transcripts yourself - they can be hundreds of thousands of lines. Run the engine and work from
13	its output.
14
15	This skill is harness-portable (agentskills.io format). Where it names a specific tool
16	(e.g. `AskUserQuestion`), treat that as "use your harness's equivalent; fall back to plain chat."
17
18	Voice: when narrating to the user, talk about mining their lore and skill candidates -
19	e.g. "I'll mine your lore here in <project> for skill candidates." Reserve the word forge for
20	the final act only: creating the skills the user selected (Step 4). Never describe mining, judging,
21	or candidates as "forge-…" anything.
22
23	## OUTPUT CONTRACT - three LAWS, read before emitting anything to the user
24
25	Named failure mode #1 (2026-06-09, BearClaude run): the agent indexed, deep-mined four skills,
26	then jumped straight to AskUserQuestion with bare option labels - ZERO dossiers rendered in chat. The
27	user declined every question because they had nothing to judge by. The entire mining run was wasted.
28
29	Named failure mode #2 (2026-06-09, BearClaude run, SAME DAY, fresh session, LAWs in effect): the
30	agent narrated phases correctly, did verification reads, then asked again with NO dossier message -
31	its last message before the question was process narration ("CodeMirrorBundle is alive in today's
32	repo…") - and the question text falsely claimed "dossiers above". Lesson: a felt self-check is not
33	a check. Compliance must be MECHANICAL: the sentinel line below is the check, not your impression.
34
35	Named failure mode #3 (2026-06-10, teammate's machine, Opus 4.8, plan-mode path): the agent DID
36	use plan-mode curation but presented a THIN plan - skill names and skip reasons with the dossiers
37	summarized away - so the user approved a forge they never saw the evidence for. The plan UI makes
38	skipping the display step impossible, not skipping the content. Lesson: the plan body must BE
39	the engine's `plan render` artifact (Step 3), which embeds every card verbatim and ends with the
40	LAW 1 sentinel. In Claude Code this is now HOOK-ENFORCED: a PreToolUse hook in this skill's
41	frontmatter denies any `ExitPlanMode` whose plan is not that artifact.
42
43	LAW 1 - DOSSIERS BEFORE CANDIDATE QUESTIONS, PROVEN BY SENTINEL. This law governs **candidate
44	decisions** - any prompt where the user chooses which skills to forge, skip, or update. (Navigation
45	questions like the Step 0.25 guided start, or scope confirmations, are exempt - they decide nothing
46	about candidates.) Before any candidate prompt you must emit one chat message that contains a full
47	dossier block (`### <name> …`, per Step 3) for EVERY candidate, and that message must END with this
48	exact line:
49
50	```
51	=== dossiers above: N ===
52	```
53
54	where N equals the number of candidates you are about to offer. At the moment of asking, the check is
55	mechanical: *"Does a prior message of mine end with `=== dossiers above: N ===` and does N match my
56	option count?"* No sentinel → you have not rendered dossiers, whatever you remember - STOP and write
57	them. Process narration between tool calls does NOT count; interim notes do NOT count.
58
59	**The strongest form of LAW 1 is plan-style curation (Claude Code, see Step 3): present the dossiers
60	AS the plan via `ExitPlanMode` - then showing the evidence and asking for the decision are the same
61	act, and skipping the display is structurally impossib

People who liked this also viewed

6 picks

Type

lore

68installs80forks

View on GitHub

How to install lore?

specstoryai/getspecstory/lore

$npx -y skills add specstoryai/getspecstory --skill lore

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Lore
2
3	Your sessions are your lore. This skill turns a user's real coding history into installed skills.
4	A deterministic engine (`scripts/mine-skills.mjs`) parses their SpecStory transcripts - from **every
5	agent** SpecStory captures (Claude Code, Codex CLI, Cursor CLI, Gemini CLI, Factory Droid, DeepSeek,
6	Antigravity, ...) - into a persistent corpus of beats, and returns corroborated candidates.
7	You - the calling agent, whichever harness you are - supply all judgment: name them, discard the
8	generic ones, and interactively forge the good ones into `SKILL.md` packages grounded in the user's
9	own commands.
10
11	The engine does the retrieval and counting; you do the synthesis. Do not try to read raw
12	transcripts yourself - they can be hundreds of thousands of lines. Run the engine and work from
13	its output.
14
15	This skill is harness-portable (agentskills.io format). Where it names a specific tool
16	(e.g. `AskUserQuestion`), treat that as "use your harness's equivalent; fall back to plain chat."
17
18	Voice: when narrating to the user, talk about mining their lore and skill candidates -
19	e.g. "I'll mine your lore here in <project> for skill candidates." Reserve the word forge for
20	the final act only: creating the skills the user selected (Step 4). Never describe mining, judging,
21	or candidates as "forge-…" anything.
22
23	## OUTPUT CONTRACT - three LAWS, read before emitting anything to the user
24
25	Named failure mode #1 (2026-06-09, BearClaude run): the agent indexed, deep-mined four skills,
26	then jumped straight to AskUserQuestion with bare option labels - ZERO dossiers rendered in chat. The
27	user declined every question because they had nothing to judge by. The entire mining run was wasted.
28
29	Named failure mode #2 (2026-06-09, BearClaude run, SAME DAY, fresh session, LAWs in effect): the
30	agent narrated phases correctly, did verification reads, then asked again with NO dossier message -
31	its last message before the question was process narration ("CodeMirrorBundle is alive in today's
32	repo…") - and the question text falsely claimed "dossiers above". Lesson: a felt self-check is not
33	a check. Compliance must be MECHANICAL: the sentinel line below is the check, not your impression.
34
35	Named failure mode #3 (2026-06-10, teammate's machine, Opus 4.8, plan-mode path): the agent DID
36	use plan-mode curation but presented a THIN plan - skill names and skip reasons with the dossiers
37	summarized away - so the user approved a forge they never saw the evidence for. The plan UI makes
38	skipping the display step impossible, not skipping the content. Lesson: the plan body must BE
39	the engine's `plan render` artifact (Step 3), which embeds every card verbatim and ends with the
40	LAW 1 sentinel. In Claude Code this is now HOOK-ENFORCED: a PreToolUse hook in this skill's
41	frontmatter denies any `ExitPlanMode` whose plan is not that artifact.
42
43	LAW 1 - DOSSIERS BEFORE CANDIDATE QUESTIONS, PROVEN BY SENTINEL. This law governs **candidate
44	decisions** - any prompt where the user chooses which skills to forge, skip, or update. (Navigation
45	questions like the Step 0.25 guided start, or scope confirmations, are exempt - they decide nothing
46	about candidates.) Before any candidate prompt you must emit one chat message that contains a full
47	dossier block (`### <name> …`, per Step 3) for EVERY candidate, and that message must END with this
48	exact line:
49
50	```
51	=== dossiers above: N ===
52	```
53
54	where N equals the number of candidates you are about to offer. At the moment of asking, the check is
55	mechanical: *"Does a prior message of mine end with `=== dossiers above: N ===` and does N match my
56	option count?"* No sentinel → you have not rendered dossiers, whatever you remember - STOP and write
57	them. Process narration between tool calls does NOT count; interim notes do NOT count.
58
59	**The strongest form of LAW 1 is plan-style curation (Claude Code, see Step 3): present the dossiers
60	AS the plan via `ExitPlanMode` - then showing the evidence and asking for the decision are the same
61	act, and skipping the display is structurally impossib

People who liked this also viewed

6 picks

Type